diff --git a/packages/ai/test/openai-compat-policy.test.ts b/packages/ai/test/openai-compat-policy.test.ts index 1a95c8aa2..1efed7c0c 100644 --- a/packages/ai/test/openai-compat-policy.test.ts +++ b/packages/ai/test/openai-compat-policy.test.ts @@ -9,6 +9,7 @@ import { import type { Model, ModelSpec, OpenAICompat } from "@oh-my-pi/pi-ai/types"; import { buildModel } from "@oh-my-pi/pi-catalog/build"; import { Effort } from "@oh-my-pi/pi-catalog/effort"; +import { getBundledModel } from "@oh-my-pi/pi-catalog/models"; function chatModel(compat: OpenAICompat): Model<"openai-completions"> { return buildModel({ @@ -160,4 +161,18 @@ describe("OpenAI compat policy", () => { expect(chatPolicy.stream.reasoningDeltasMayBeCumulative).toBe(true); expect(responsesPolicy.stream.reasoningDeltasMayBeCumulative).toBe(true); }); + + it("routes Token Plan qwen3.8-max effort selections onto the wire", () => { + const model = getBundledModel<"openai-completions">("alibaba-token-plan", "qwen3.8-max"); + for (const effort of [Effort.Low, Effort.Medium, Effort.XHigh]) { + const params = chatParams(); + applyChatCompletionsCompatPolicy( + params, + resolveOpenAICompatPolicy(model, { endpoint: "chat-completions", reasoning: effort }), + ); + expect(params.reasoning_effort).toBe(effort); + expect(params.enable_thinking).toBeUndefined(); + expect(params.chat_template_kwargs).toBeUndefined(); + } + }); }); diff --git a/packages/catalog/CHANGELOG.md b/packages/catalog/CHANGELOG.md index fa274afdd..6cfbc88e2 100644 --- a/packages/catalog/CHANGELOG.md +++ b/packages/catalog/CHANGELOG.md @@ -2,6 +2,10 @@ ## [Unreleased] +### Fixed + +- Fixed dynamically discovered `alibaba-token-plan/qwen3.8-max` metadata so thinking controls and image input are available ([#8019](https://github.com/can1357/oh-my-pi/issues/8019)). + ## [17.2.11] - 2026-08-07 ### Fixed diff --git a/packages/catalog/scripts/generated-policies.ts b/packages/catalog/scripts/generated-policies.ts index b68004971..b53631849 100644 --- a/packages/catalog/scripts/generated-policies.ts +++ b/packages/catalog/scripts/generated-policies.ts @@ -172,7 +172,13 @@ export function applyGeneratedModelPolicies(models: ModelSpec[]): void { */ export function rebakeModelThinking(model: ModelSpec): void { if (isVariantCollapsedSpec(model)) return; - if (model.provider === "alibaba-token-plan" && model.id === "qwen3.8-max-preview" && model.thinking) return; + if ( + model.provider === "alibaba-token-plan" && + (model.id === "qwen3.8-max-preview" || model.id === "qwen3.8-max") && + model.thinking + ) { + return; + } const requiresProviderAuthoredEffort = model.provider === "umans" && (model.thinking?.requiresEffort === true || model.id === "umans-kimi-k2.7"); const thinking = resolveModelThinking({ ...model, thinking: undefined }, buildCompat(model)); diff --git a/packages/catalog/src/models.json b/packages/catalog/src/models.json index 99910f606..3978d0ba4 100644 --- a/packages/catalog/src/models.json +++ b/packages/catalog/src/models.json @@ -16,16 +16,16 @@ "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 1048576, + "contextWindow": 1000000, "maxTokens": 393216, "thinking": { "mode": "effort", "efforts": [ + "low", "high", "max" ] - }, - "supportsComputerUse": false + } }, "deepseek-ai/deepseek-v4-pro": { "id": "deepseek-ai/deepseek-v4-pro", @@ -43,7 +43,7 @@ "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 1048576, + "contextWindow": 1000000, "maxTokens": 393216, "thinking": { "mode": "effort", @@ -51,12 +51,11 @@ "high", "max" ] - }, - "supportsComputerUse": false + } }, "google/gemma-4-31b-it": { "id": "google/gemma-4-31b-it", - "name": "Gemma 4 31B", + "name": "Gemma 4 31B IT", "api": "openai-completions", "provider": "aiand", "baseUrl": "https://api.aiand.com/v1", @@ -72,7 +71,7 @@ "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 131072, + "maxTokens": 262144, "thinking": { "mode": "effort", "efforts": [ @@ -82,8 +81,37 @@ "high", "xhigh" ] + } + }, + "moonshotai/kimi-k2.6": { + "id": "moonshotai/kimi-k2.6", + "name": "Kimi K2.6", + "api": "openai-completions", + "provider": "aiand", + "baseUrl": "https://api.aiand.com/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.85, + "output": 3.5, + "cacheRead": 0, + "cacheWrite": 0 }, - "supportsComputerUse": false + "contextWindow": 262144, + "maxTokens": 262144, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "moonshotai/kimi-k2.7-code": { "id": "moonshotai/kimi-k2.7-code", @@ -113,8 +141,7 @@ "high", "xhigh" ] - }, - "supportsComputerUse": false + } }, "moonshotai/kimi-k3": { "id": "moonshotai/kimi-k3", @@ -175,8 +202,7 @@ "medium", "high" ] - }, - "supportsComputerUse": false + } }, "qwen/qwen3.6-27b": { "id": "qwen/qwen3.6-27b", @@ -190,13 +216,13 @@ "image" ], "cost": { - "input": 0.32, - "output": 3.2, + "input": 0, + "output": 0, "cacheRead": 0, "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 65536, + "maxTokens": 262144, "thinking": { "mode": "effort", "efforts": [ @@ -205,8 +231,36 @@ "medium", "high" ] + } + }, + "zai-org/glm-5.1": { + "id": "zai-org/glm-5.1", + "name": "GLM 5.1", + "api": "openai-completions", + "provider": "aiand", + "baseUrl": "https://api.aiand.com/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 1.4, + "output": 4.4, + "cacheRead": 0, + "cacheWrite": 0 }, - "supportsComputerUse": false + "contextWindow": 202752, + "maxTokens": 131072, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "zai-org/glm-5.2": { "id": "zai-org/glm-5.2", @@ -224,7 +278,7 @@ "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 1048576, + "contextWindow": 1000000, "maxTokens": 131072, "thinking": { "mode": "effort", @@ -235,8 +289,7 @@ "high", "max" ] - }, - "supportsComputerUse": false + } } }, "aimlapi": { @@ -2266,11 +2319,11 @@ }, "claude-opus-4-1-20250805": { "id": "claude-opus-4-1-20250805", - "name": "Claude Opus 4.1", + "name": "Claude 4.1 Opus", "api": "openai-completions", "provider": "aimlapi", "baseUrl": "https://api.aimlapi.com/v1", - "reasoning": true, + "reasoning": false, "input": [ "text", "image" @@ -2282,17 +2335,7 @@ "cacheWrite": 0 }, "contextWindow": 200000, - "maxTokens": 32000, - "thinking": { - "mode": "effort", - "efforts": [ - "minimal", - "low", - "medium", - "high", - "xhigh" - ] - } + "maxTokens": 32000 }, "claude-opus-4-20250514": { "id": "claude-opus-4-20250514", @@ -2649,6 +2692,7 @@ "thinking": { "mode": "effort", "efforts": [ + "low", "high", "max" ] @@ -2727,7 +2771,7 @@ }, "deepseek/deepseek-r1": { "id": "deepseek/deepseek-r1", - "name": "R1", + "name": "DeepSeek-R1", "api": "openai-completions", "provider": "aimlapi", "baseUrl": "https://api.aimlapi.com/v1", @@ -5182,7 +5226,7 @@ }, "minimax/minimax-m3": { "id": "minimax/minimax-m3", - "name": "MiniMax M3", + "name": "MiniMax-M3", "api": "openai-completions", "provider": "aimlapi", "baseUrl": "https://api.aimlapi.com/v1", @@ -7427,7 +7471,7 @@ }, "xiaomi/mimo-v2.5-pro": { "id": "xiaomi/mimo-v2.5-pro", - "name": "MiMo V2.5 Pro", + "name": "MiMo-V2.5-Pro", "api": "openai-completions", "provider": "aimlapi", "baseUrl": "https://api.aimlapi.com/v1", @@ -7962,6 +8006,40 @@ "supportsDeveloperRole": false } }, + "qwen3.8-max": { + "id": "qwen3.8-max", + "name": "Qwen3.8 Max", + "api": "openai-completions", + "provider": "alibaba-token-plan", + "baseUrl": "https://token-plan.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0, + "output": 0, + "cacheRead": 0, + "cacheWrite": 0 + }, + "contextWindow": 1000000, + "maxTokens": 131072, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "xhigh" + ], + "defaultLevel": "xhigh" + }, + "compat": { + "supportsDeveloperRole": false, + "supportsReasoningEffort": true, + "thinkingFormat": "openai" + } + }, "qwen3.8-max-preview": { "id": "qwen3.8-max-preview", "name": "Qwen3.8 Max Preview", @@ -7970,8 +8048,7 @@ "baseUrl": "https://token-plan.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1", "reasoning": true, "input": [ - "text", - "image" + "text" ], "cost": { "input": 0, @@ -7992,7 +8069,8 @@ }, "compat": { "supportsDeveloperRole": false, - "supportsReasoningEffort": true + "supportsReasoningEffort": true, + "thinkingFormat": "openai" } } }, @@ -11942,8 +12020,7 @@ "max" ], "supportsDisplay": true - }, - "supportsComputerUse": false + } }, "claude-opus-4-0": { "id": "claude-opus-4-0", @@ -12009,11 +12086,11 @@ }, "claude-opus-4-1-20250805": { "id": "claude-opus-4-1-20250805", - "name": "Claude Opus 4.1", + "name": "Claude 4.1 Opus", "api": "anthropic-messages", "provider": "anthropic", "baseUrl": "https://api.anthropic.com", - "reasoning": true, + "reasoning": false, "input": [ "text", "image" @@ -12025,17 +12102,7 @@ "cacheWrite": 18.75 }, "contextWindow": 200000, - "maxTokens": 32000, - "thinking": { - "mode": "budget", - "efforts": [ - "minimal", - "low", - "medium", - "high", - "xhigh" - ] - } + "maxTokens": 32000 }, "claude-opus-4-20250514": { "id": "claude-opus-4-20250514", @@ -13214,10 +13281,10 @@ "image" ], "cost": { - "input": 0.2, - "output": 1.2, - "cacheRead": 0.02, - "cacheWrite": 0 + "input": 1, + "output": 6, + "cacheRead": 0.1, + "cacheWrite": 1.25 }, "contextWindow": 1050000, "maxTokens": 128000, @@ -13274,10 +13341,10 @@ "image" ], "cost": { - "input": 2, - "output": 12, - "cacheRead": 0.2, - "cacheWrite": 0 + "input": 2.5, + "output": 15, + "cacheRead": 0.25, + "cacheWrite": 3.125 }, "contextWindow": 1050000, "maxTokens": 128000, @@ -15453,6 +15520,34 @@ "thinking": { "mode": "effort", "efforts": [ + "low", + "high", + "max" + ] + } + }, + "deepseek-ai/DeepSeek-V4-Flash-0731": { + "id": "deepseek-ai/DeepSeek-V4-Flash-0731", + "name": "DeepSeek V4 Flash 0731", + "api": "openai-completions", + "provider": "coreweave", + "baseUrl": "https://api.inference.wandb.ai/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.13, + "output": 0.28, + "cacheRead": 0.07, + "cacheWrite": 0 + }, + "contextWindow": 262144, + "maxTokens": 262144, + "thinking": { + "mode": "effort", + "efforts": [ + "low", "high", "max" ] @@ -15469,9 +15564,9 @@ "text" ], "cost": { - "input": 1.74, - "output": 3.48, - "cacheRead": 0.14, + "input": 1.15, + "output": 2.55, + "cacheRead": 0.2, "cacheWrite": 0 }, "contextWindow": 1048576, @@ -15962,7 +16057,7 @@ }, "Qwen/Qwen3-235B-A22B-Instruct-2507": { "id": "Qwen/Qwen3-235B-A22B-Instruct-2507", - "name": "Qwen3 235B A22B-2507", + "name": "Qwen3 235B-A22B Instruct 2507", "api": "openai-completions", "provider": "coreweave", "baseUrl": "https://api.inference.wandb.ai/v1", @@ -15981,7 +16076,7 @@ }, "Qwen/Qwen3-235B-A22B-Thinking-2507": { "id": "Qwen/Qwen3-235B-A22B-Thinking-2507", - "name": "Qwen3 235B A22B Thinking-2507", + "name": "Qwen3-235B-A22B-Thinking-2507", "api": "openai-completions", "provider": "coreweave", "baseUrl": "https://api.inference.wandb.ai/v1", @@ -16048,7 +16143,7 @@ }, "Qwen/Qwen3.5-27B": { "id": "Qwen/Qwen3.5-27B", - "name": "Qwen3.5-27B", + "name": "Qwen3.5 27B", "api": "openai-completions", "provider": "coreweave", "baseUrl": "https://api.inference.wandb.ai/v1", @@ -20939,6 +21034,7 @@ "thinking": { "mode": "effort", "efforts": [ + "low", "high", "max" ] @@ -21063,6 +21159,7 @@ "thinking": { "mode": "effort", "efforts": [ + "low", "high", "max" ] @@ -21094,6 +21191,7 @@ "thinking": { "mode": "effort", "efforts": [ + "low", "high", "max" ] @@ -22430,6 +22528,45 @@ "requiresEffort": true } }, + "gemini-3.6-flash": { + "id": "gemini-3.6-flash", + "name": "Gemini 3.6 Flash", + "api": "openai-completions", + "provider": "github-copilot", + "baseUrl": "https://api.githubcopilot.com", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 1.5, + "output": 7.5, + "cacheRead": 0.15, + "cacheWrite": 0 + }, + "contextWindow": 1000000, + "maxTokens": 64000, + "headers": { + "User-Agent": "opencode/1.3.15", + "X-GitHub-Api-Version": "2026-06-01" + }, + "compat": { + "supportsStore": false, + "supportsDeveloperRole": false, + "supportsReasoningEffort": false + }, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ], + "requiresEffort": true + } + }, "gpt-4.1": { "id": "gpt-4.1", "name": "GPT-4.1", @@ -22923,10 +23060,10 @@ "image" ], "cost": { - "input": 1, - "output": 6, - "cacheRead": 0.1, - "cacheWrite": 1.25 + "input": 0.2, + "output": 1.2, + "cacheRead": 0.02, + "cacheWrite": 0 }, "contextWindow": 1050000, "maxTokens": 128000, @@ -22991,10 +23128,10 @@ "image" ], "cost": { - "input": 2.5, - "output": 15, - "cacheRead": 0.25, - "cacheWrite": 3.125 + "input": 2, + "output": 12, + "cacheRead": 0.2, + "cacheWrite": 0 }, "contextWindow": 1050000, "maxTokens": 128000, @@ -23013,6 +23150,40 @@ ] } }, + "grok-4.5": { + "id": "grok-4.5", + "name": "Grok 4.5", + "api": "openai-responses", + "provider": "github-copilot", + "baseUrl": "https://api.githubcopilot.com", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 2, + "output": 6, + "cacheRead": 0.5, + "cacheWrite": 0 + }, + "contextWindow": 500000, + "maxTokens": 128000, + "headers": { + "User-Agent": "opencode/1.3.15", + "X-GitHub-Api-Version": "2026-06-01" + }, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } + }, "grok-code-fast-1": { "id": "grok-code-fast-1", "name": "Grok Code Fast 1", @@ -23087,6 +23258,48 @@ ] } }, + "kimi-k3": { + "id": "kimi-k3", + "name": "Kimi K3", + "api": "openai-completions", + "provider": "github-copilot", + "baseUrl": "https://api.githubcopilot.com", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 3, + "output": 15, + "cacheRead": 0.3, + "cacheWrite": 0 + }, + "contextWindow": 1048576, + "maxTokens": 131072, + "headers": { + "User-Agent": "opencode/1.3.15", + "X-GitHub-Api-Version": "2026-06-01" + }, + "compat": { + "supportsStore": false, + "supportsDeveloperRole": false, + "supportsReasoningEffort": false + }, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "high", + "max" + ], + "defaultLevel": "max", + "effortMap": { + "max": "max" + }, + "requiresEffort": true + } + }, "mai-code-1-flash-picker": { "id": "mai-code-1-flash-picker", "name": "MAI-Code-1-Flash", @@ -23686,6 +23899,7 @@ "thinking": { "mode": "effort", "efforts": [ + "low", "high", "max" ] @@ -26672,7 +26886,7 @@ }, "qwen/qwen3-32b": { "id": "qwen/qwen3-32b", - "name": "Qwen3-32B", + "name": "Qwen3 32B", "api": "openai-completions", "provider": "groq", "baseUrl": "https://api.groq.com/openai/v1", @@ -26703,6 +26917,35 @@ "high": "default" } } + }, + "qwen/qwen3.6-27b": { + "id": "qwen/qwen3.6-27b", + "name": "Qwen3.6 27B", + "api": "openai-completions", + "provider": "groq", + "baseUrl": "https://api.groq.com/openai/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.6, + "output": 3, + "cacheRead": 0.3, + "cacheWrite": 0 + }, + "contextWindow": 131072, + "maxTokens": 16384, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + } } }, "huggingface": { @@ -26758,9 +27001,9 @@ ] } }, - "deepseek-ai/DeepSeek-V3.1": { - "id": "deepseek-ai/DeepSeek-V3.1", - "name": "DeepSeek V3.1", + "deepseek-ai/DeepSeek-V3": { + "id": "deepseek-ai/DeepSeek-V3", + "name": "DeepSeek-V3", "api": "openai-completions", "provider": "huggingface", "baseUrl": "https://router.huggingface.co/v1", @@ -26769,13 +27012,39 @@ "text" ], "cost": { - "input": 0.6, - "output": 1.25, - "cacheRead": 0.6, - "cacheWrite": 0.6 + "input": 0.4, + "output": 1.3, + "cacheRead": 0, + "cacheWrite": 0 + }, + "contextWindow": 64000, + "maxTokens": 8192 + }, + "deepseek-ai/DeepSeek-V3.1": { + "id": "deepseek-ai/DeepSeek-V3.1", + "name": "DeepSeek-V3.1", + "api": "openai-completions", + "provider": "huggingface", + "baseUrl": "https://router.huggingface.co/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.27, + "output": 1, + "cacheRead": 0, + "cacheWrite": 0 }, "contextWindow": 131072, - "maxTokens": 8192 + "maxTokens": 8192, + "thinking": { + "mode": "effort", + "efforts": [ + "high", + "max" + ] + } }, "deepseek-ai/DeepSeek-V3.2": { "id": "deepseek-ai/DeepSeek-V3.2", @@ -26824,6 +27093,34 @@ "thinking": { "mode": "effort", "efforts": [ + "low", + "high", + "max" + ] + } + }, + "deepseek-ai/DeepSeek-V4-Flash-0731": { + "id": "deepseek-ai/DeepSeek-V4-Flash-0731", + "name": "DeepSeek V4 Flash 0731", + "api": "openai-completions", + "provider": "huggingface", + "baseUrl": "https://router.huggingface.co/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.14, + "output": 0.28, + "cacheRead": 0, + "cacheWrite": 0 + }, + "contextWindow": 1048576, + "maxTokens": 384000, + "thinking": { + "mode": "effort", + "efforts": [ + "low", "high", "max" ] @@ -27368,6 +27665,25 @@ ] } }, + "Qwen/Qwen3-235B-A22B-Instruct-2507": { + "id": "Qwen/Qwen3-235B-A22B-Instruct-2507", + "name": "Qwen3 235B-A22B Instruct 2507", + "api": "openai-completions", + "provider": "huggingface", + "baseUrl": "https://router.huggingface.co/v1", + "reasoning": false, + "input": [ + "text" + ], + "cost": { + "input": 0.855, + "output": 2.565, + "cacheRead": 0, + "cacheWrite": 0 + }, + "contextWindow": 262144, + "maxTokens": 16384 + }, "Qwen/Qwen3-235B-A22B-Thinking-2507": { "id": "Qwen/Qwen3-235B-A22B-Thinking-2507", "name": "Qwen3-235B-A22B-Thinking-2507", @@ -27841,6 +28157,36 @@ ] } }, + "thinkingmachines/Inkling-Small": { + "id": "thinkingmachines/Inkling-Small", + "name": "Inkling Small", + "api": "openai-completions", + "provider": "huggingface", + "baseUrl": "https://router.huggingface.co/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.5, + "output": 1.2, + "cacheRead": 0, + "cacheWrite": 0 + }, + "contextWindow": 524288, + "maxTokens": 1048576, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } + }, "XiaomiMiMo/MiMo-V2-Flash": { "id": "XiaomiMiMo/MiMo-V2-Flash", "name": "MiMo-V2-Flash", @@ -28192,23 +28538,33 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 10, + "output": 50, + "cacheRead": 1, + "cacheWrite": 12.5 }, - "contextWindow": null, - "maxTokens": null, - "supportsComputerUse": false + "contextWindow": 1000000, + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "~anthropic/claude-haiku-latest": { "id": "~anthropic/claude-haiku-latest", - "name": "Claude Haiku Latest", + "name": "Anthropic Claude Haiku Latest", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -28268,7 +28624,7 @@ }, "~anthropic/claude-sonnet-latest": { "id": "~anthropic/claude-sonnet-latest", - "name": "Claude Sonnet Latest", + "name": "Anthropic Claude Sonnet Latest", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -28278,10 +28634,10 @@ "image" ], "cost": { - "input": 3, - "output": 15, - "cacheRead": 0.3, - "cacheWrite": 3.75 + "input": 2, + "output": 10, + "cacheRead": 0.2, + "cacheWrite": 2.5 }, "contextWindow": 1000000, "maxTokens": 128000, @@ -28302,23 +28658,30 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.09, + "output": 0.18, + "cacheRead": 0.018, "cacheWrite": 0 }, "contextWindow": 1048576, - "maxTokens": 1048576, - "supportsComputerUse": false + "maxTokens": 384000, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "high", + "max" + ] + } }, "~google/gemini-flash-latest": { "id": "~google/gemini-flash-latest", - "name": "Gemini Flash Latest", + "name": "Google Gemini Flash Latest", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -28329,9 +28692,9 @@ ], "cost": { "input": 1.5, - "output": 9, + "output": 7.5, "cacheRead": 0.15, - "cacheWrite": 0.08333333333333334 + "cacheWrite": 0.083333 }, "contextWindow": 1048576, "maxTokens": 65536, @@ -28348,7 +28711,7 @@ }, "~google/gemini-pro-latest": { "id": "~google/gemini-pro-latest", - "name": "Gemini Pro Latest", + "name": "Google Gemini Pro Latest", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -28378,7 +28741,7 @@ }, "~moonshotai/kimi-latest": { "id": "~moonshotai/kimi-latest", - "name": "Kimi Latest", + "name": "MoonshotAI Kimi Latest", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -28388,13 +28751,13 @@ "image" ], "cost": { - "input": 0.74, - "output": 3.49, - "cacheRead": 0.14, + "input": 2.5, + "output": 14, + "cacheRead": 0.29, "cacheWrite": 0 }, - "contextWindow": 262142, - "maxTokens": 262142, + "contextWindow": 1048576, + "maxTokens": 1048576, "thinking": { "mode": "effort", "efforts": [ @@ -28408,7 +28771,7 @@ }, "~openai/gpt-latest": { "id": "~openai/gpt-latest", - "name": "GPT Latest", + "name": "OpenAI GPT Latest", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -28421,7 +28784,7 @@ "input": 5, "output": 30, "cacheRead": 0.5, - "cacheWrite": 0 + "cacheWrite": 6.25 }, "contextWindow": 1050000, "maxTokens": 128000, @@ -28438,7 +28801,7 @@ }, "~openai/gpt-mini-latest": { "id": "~openai/gpt-mini-latest", - "name": "GPT Mini Latest", + "name": "OpenAI GPT Mini Latest", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -28472,19 +28835,29 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 2, + "output": 6, + "cacheRead": 0.3, "cacheWrite": 0 }, - "contextWindow": null, - "maxTokens": null, - "supportsComputerUse": false + "contextWindow": 500000, + "maxTokens": 500000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "ai21/jamba-large-1.7": { "id": "ai21/jamba-large-1.7", @@ -28553,19 +28926,28 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.8, + "output": 1.6, + "cacheRead": 0.2, "cacheWrite": 0 }, - "contextWindow": null, - "maxTokens": null, - "supportsComputerUse": false + "contextWindow": 131072, + "maxTokens": 32768, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "aion-labs/aion-3.0": { "id": "aion-labs/aion-3.0", @@ -28573,19 +28955,28 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 3, + "output": 6, + "cacheRead": 0.75, "cacheWrite": 0 }, - "contextWindow": null, - "maxTokens": null, - "supportsComputerUse": false + "contextWindow": 131072, + "maxTokens": 32768, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "aion-labs/aion-3.0-mini": { "id": "aion-labs/aion-3.0-mini", @@ -28593,19 +28984,28 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.7, + "output": 1.4, + "cacheRead": 0.18, "cacheWrite": 0 }, - "contextWindow": null, - "maxTokens": null, - "supportsComputerUse": false + "contextWindow": 131072, + "maxTokens": 32768, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "aion-labs/aion-rp-llama-3.1-8b": { "id": "aion-labs/aion-rp-llama-3.1-8b", @@ -28919,7 +29319,7 @@ "cost": { "input": 2.5, "output": 12.5, - "cacheRead": 0, + "cacheRead": 0.625, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -29003,7 +29403,8 @@ "cacheWrite": 1 }, "contextWindow": 200000, - "maxTokens": 8192 + "maxTokens": 8192, + "supportsComputerUse": false }, "anthropic/claude-3.5-sonnet": { "id": "anthropic/claude-3.5-sonnet", @@ -29092,10 +29493,10 @@ "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 10, + "output": 50, + "cacheRead": 1, + "cacheWrite": 12.5 }, "contextWindow": 1000000, "maxTokens": 128000, @@ -29108,8 +29509,7 @@ "high", "xhigh" ] - }, - "supportsComputerUse": false + } }, "anthropic/claude-haiku-4.5": { "id": "anthropic/claude-haiku-4.5", @@ -29289,7 +29689,8 @@ "high", "xhigh" ] - } + }, + "supportsComputerUse": false }, "anthropic/claude-opus-4.7": { "id": "anthropic/claude-opus-4.7", @@ -29323,7 +29724,7 @@ }, "anthropic/claude-opus-4.7-fast": { "id": "anthropic/claude-opus-4.7-fast", - "name": "Claude Opus 4.7 (Fast)", + "name": "Claude Opus 4.7", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -29363,10 +29764,10 @@ "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 5, + "output": 25, + "cacheRead": 0.5, + "cacheWrite": 6.25 }, "contextWindow": 1000000, "maxTokens": 128000, @@ -29379,28 +29780,37 @@ "high", "xhigh" ] - }, - "supportsComputerUse": false + } }, "anthropic/claude-opus-4.8-fast": { "id": "anthropic/claude-opus-4.8-fast", - "name": "Claude Opus 4.8 (Fast)", + "name": "Claude Opus 4.8", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 10, + "output": 50, + "cacheRead": 1, + "cacheWrite": 12.5 }, "contextWindow": 1000000, "maxTokens": 128000, - "supportsComputerUse": false + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "anthropic/claude-opus-5": { "id": "anthropic/claude-opus-5", @@ -29414,14 +29824,13 @@ "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 5, + "output": 25, + "cacheRead": 0.5, + "cacheWrite": 6.25 }, "contextWindow": 1000000, "maxTokens": 128000, - "supportsComputerUse": false, "thinking": { "mode": "effort", "efforts": [ @@ -29435,23 +29844,33 @@ }, "anthropic/claude-opus-5-fast": { "id": "anthropic/claude-opus-5-fast", - "name": "Claude Opus 5 (Fast)", + "name": "Claude Opus 5", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 10, + "output": 50, + "cacheRead": 1, + "cacheWrite": 12.5 }, "contextWindow": 1000000, "maxTokens": 128000, - "supportsComputerUse": false + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "anthropic/claude-sonnet-4": { "id": "anthropic/claude-sonnet-4", @@ -29527,8 +29946,8 @@ "cost": { "input": 3, "output": 15, - "cacheRead": 0, - "cacheWrite": 0 + "cacheRead": 0.3, + "cacheWrite": 3.75 }, "contextWindow": 1000000, "maxTokens": 128000, @@ -29555,10 +29974,10 @@ "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 2, + "output": 10, + "cacheRead": 0.2, + "cacheWrite": 2.5 }, "contextWindow": 1000000, "maxTokens": 128000, @@ -29571,8 +29990,7 @@ "high", "xhigh" ] - }, - "supportsComputerUse": false + } }, "arcee-ai/coder-large": { "id": "arcee-ai/coder-large", @@ -29692,7 +30110,7 @@ "cost": { "input": 0.22, "output": 0.85, - "cacheRead": 0, + "cacheRead": 0.06, "cacheWrite": 0 }, "contextWindow": 262144, @@ -29757,7 +30175,8 @@ "high", "xhigh" ] - } + }, + "supportsComputerUse": false }, "arcee-ai/virtuoso-large": { "id": "arcee-ai/virtuoso-large", @@ -29805,7 +30224,8 @@ "high", "xhigh" ] - } + }, + "supportsComputerUse": false }, "baidu/ernie-4.5-21b-a3b": { "id": "baidu/ernie-4.5-21b-a3b", @@ -29824,7 +30244,8 @@ "cacheWrite": 0 }, "contextWindow": 120000, - "maxTokens": 8000 + "maxTokens": 8000, + "supportsComputerUse": false }, "baidu/ernie-4.5-300b-a47b": { "id": "baidu/ernie-4.5-300b-a47b", @@ -29875,7 +30296,8 @@ "high", "xhigh" ] - } + }, + "supportsComputerUse": false }, "baidu/ernie-4.5-vl-424b-a47b": { "id": "baidu/ernie-4.5-vl-424b-a47b", @@ -30142,7 +30564,7 @@ }, "cohere/command-r-08-2024": { "id": "cohere/command-r-08-2024", - "name": "Command R (08-2024)", + "name": "Command R", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -30161,7 +30583,7 @@ }, "cohere/command-r-plus-08-2024": { "id": "cohere/command-r-plus-08-2024", - "name": "Command R+ (08-2024)", + "name": "Command R+", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -30195,7 +30617,8 @@ "cacheWrite": 0 }, "contextWindow": 128000, - "maxTokens": 4000 + "maxTokens": 4000, + "supportsComputerUse": false }, "cohere/north-mini-code:free": { "id": "cohere/north-mini-code:free", @@ -30203,7 +30626,7 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ "text" ], @@ -30215,7 +30638,16 @@ }, "contextWindow": 256000, "maxTokens": 64000, - "supportsComputerUse": false + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "corethink:free": { "id": "corethink:free", @@ -30260,7 +30692,7 @@ }, "deepseek/deepseek-chat": { "id": "deepseek/deepseek-chat", - "name": "DeepSeek V3", + "name": "DeepSeek Chat", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -30269,13 +30701,13 @@ "text" ], "cost": { - "input": 0.32, - "output": 0.89, - "cacheRead": 0.15, + "input": 0.4, + "output": 1.3, + "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 163840, - "maxTokens": 163840 + "contextWindow": 128000, + "maxTokens": 16000 }, "deepseek/deepseek-chat-v3-0324": { "id": "deepseek/deepseek-chat-v3-0324", @@ -30288,9 +30720,9 @@ "text" ], "cost": { - "input": 0.2, - "output": 0.77, - "cacheRead": 0.095, + "input": 0.27, + "output": 1.12, + "cacheRead": 0.135, "cacheWrite": 0 }, "contextWindow": 163840, @@ -30307,13 +30739,13 @@ "text" ], "cost": { - "input": 0.15, - "output": 0.75, - "cacheRead": 0, + "input": 0.27, + "output": 1, + "cacheRead": 0.135, "cacheWrite": 0 }, - "contextWindow": 32768, - "maxTokens": 7168, + "contextWindow": 163840, + "maxTokens": 32768, "thinking": { "mode": "effort", "efforts": [ @@ -30324,7 +30756,7 @@ }, "deepseek/deepseek-r1": { "id": "deepseek/deepseek-r1", - "name": "R1", + "name": "DeepSeek-R1", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -30359,13 +30791,13 @@ "text" ], "cost": { - "input": 0.45, - "output": 2.15, - "cacheRead": 0.2, + "input": 0.7, + "output": 2.5, + "cacheRead": 0.35, "cacheWrite": 0 }, "contextWindow": 163840, - "maxTokens": 65536, + "maxTokens": 32768, "thinking": { "mode": "effort", "efforts": [ @@ -30426,12 +30858,12 @@ "text" ], "cost": { - "input": 0.21, - "output": 0.79, - "cacheRead": 0.13, + "input": 0.27, + "output": 1, + "cacheRead": 0.135, "cacheWrite": 0 }, - "contextWindow": 163840, + "contextWindow": 131072, "maxTokens": 32768, "thinking": { "mode": "effort", @@ -30473,13 +30905,13 @@ "text" ], "cost": { - "input": 0.26, - "output": 0.38, - "cacheRead": 0.125, + "input": 0.269, + "output": 0.4, + "cacheRead": 0.1345, "cacheWrite": 0 }, "contextWindow": 163840, - "maxTokens": 65536, + "maxTokens": 163840, "thinking": { "mode": "effort", "efforts": [ @@ -30548,14 +30980,15 @@ "cost": { "input": 0.14, "output": 0.28, - "cacheRead": 0.0028, + "cacheRead": 0.028, "cacheWrite": 0 }, "contextWindow": 1048576, - "maxTokens": 384000, + "maxTokens": 393216, "thinking": { "mode": "effort", "efforts": [ + "low", "high", "max" ] @@ -30567,19 +31000,26 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.14, + "output": 0.28, + "cacheRead": 0.028, "cacheWrite": 0 }, "contextWindow": 1048576, "maxTokens": 384000, - "supportsComputerUse": false + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "high", + "max" + ] + } }, "deepseek/deepseek-v4-flash:discounted": { "id": "deepseek/deepseek-v4-flash:discounted", @@ -30587,19 +31027,26 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.14, + "output": 0.28, + "cacheRead": 0.0028, "cacheWrite": 0 }, "contextWindow": 1048576, "maxTokens": 384000, - "supportsComputerUse": false + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "high", + "max" + ] + } }, "deepseek/deepseek-v4-flash:free": { "id": "deepseek/deepseek-v4-flash:free", @@ -30633,9 +31080,9 @@ "text" ], "cost": { - "input": 0.435, - "output": 0.87, - "cacheRead": 0.003625, + "input": 1.6, + "output": 3.2, + "cacheRead": 0.135, "cacheWrite": 0 }, "contextWindow": 1048576, @@ -30654,19 +31101,25 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.435, + "output": 0.87, + "cacheRead": 0.003625, "cacheWrite": 0 }, "contextWindow": 1048576, "maxTokens": 384000, - "supportsComputerUse": false + "thinking": { + "mode": "effort", + "efforts": [ + "high", + "max" + ] + } }, "eleutherai/llemma_7b": { "id": "eleutherai/llemma_7b", @@ -30706,7 +31159,8 @@ "cacheWrite": 0 }, "contextWindow": 32768, - "maxTokens": 6554 + "maxTokens": 6554, + "supportsComputerUse": false }, "giga-potato": { "id": "giga-potato", @@ -30765,7 +31219,8 @@ "cacheWrite": 0.083333 }, "contextWindow": 1048576, - "maxTokens": 8192 + "maxTokens": 8192, + "supportsComputerUse": false }, "google/gemini-2.0-flash-lite-001": { "id": "google/gemini-2.0-flash-lite-001", @@ -30785,7 +31240,8 @@ "cacheWrite": 0 }, "contextWindow": 1048576, - "maxTokens": 8192 + "maxTokens": 8192, + "supportsComputerUse": false }, "google/gemini-2.5-flash": { "id": "google/gemini-2.5-flash", @@ -30838,7 +31294,7 @@ }, "google/gemini-2.5-flash-lite": { "id": "google/gemini-2.5-flash-lite", - "name": "Gemini 2.5 Flash Lite", + "name": "Gemini 2.5 Flash-Lite", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -30892,7 +31348,8 @@ "medium", "high" ] - } + }, + "supportsComputerUse": false }, "google/gemini-2.5-pro": { "id": "google/gemini-2.5-pro", @@ -31016,23 +31473,31 @@ }, "google/gemini-3-pro-image": { "id": "google/gemini-3-pro-image", - "name": "Nano Banana Pro (Gemini 3 Pro Image)", + "name": "Nano Banana Pro", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 2, + "output": 12, + "cacheRead": 0.2, + "cacheWrite": 0.375 }, "contextWindow": 65536, "maxTokens": 32768, - "supportsComputerUse": false + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "high" + ], + "requiresEffort": true + } }, "google/gemini-3-pro-image-preview": { "id": "google/gemini-3-pro-image-preview", @@ -31148,7 +31613,7 @@ "input": 0.25, "output": 1.5, "cacheRead": 0.025, - "cacheWrite": 0.08333 + "cacheWrite": 0.083333 }, "contextWindow": 1048576, "maxTokens": 65536, @@ -31197,8 +31662,8 @@ "cost": { "input": 0.25, "output": 1.5, - "cacheRead": 0, - "cacheWrite": 0 + "cacheRead": 0.025, + "cacheWrite": 0.083333 }, "contextWindow": 1048576, "maxTokens": 65536, @@ -31227,8 +31692,8 @@ "cost": { "input": 2, "output": 12, - "cacheRead": 0, - "cacheWrite": 0 + "cacheRead": 0.2, + "cacheWrite": 0.375 }, "contextWindow": 1048576, "maxTokens": 65536, @@ -31255,8 +31720,8 @@ "cost": { "input": 2, "output": 12, - "cacheRead": 0, - "cacheWrite": 0 + "cacheRead": 0.2, + "cacheWrite": 0.375 }, "contextWindow": 1048576, "maxTokens": 65536, @@ -31284,7 +31749,7 @@ "input": 1.5, "output": 9, "cacheRead": 0.15, - "cacheWrite": 0.08333 + "cacheWrite": 0.083333 }, "contextWindow": 1048576, "maxTokens": 65536, @@ -31305,19 +31770,29 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 0.3, + "output": 2.5, + "cacheRead": 0.03, + "cacheWrite": 0.083333 }, "contextWindow": 1048576, "maxTokens": 65536, - "supportsComputerUse": false + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ], + "requiresEffort": true + } }, "google/gemini-3.6-flash": { "id": "google/gemini-3.6-flash", @@ -31325,19 +31800,29 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 1.5, + "output": 7.5, + "cacheRead": 0.15, + "cacheWrite": 0.083333 }, "contextWindow": 1048576, "maxTokens": 65536, - "supportsComputerUse": false + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ], + "requiresEffort": true + } }, "google/gemma-2-27b-it": { "id": "google/gemma-2-27b-it", @@ -31382,7 +31867,7 @@ }, "google/gemma-3-12b-it": { "id": "google/gemma-3-12b-it", - "name": "Gemma 3 12B IT", + "name": "Gemma 3 12B", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -31392,14 +31877,13 @@ "image" ], "cost": { - "input": 0, - "output": 0, + "input": 0.05, + "output": 0.15, "cacheRead": 0, "cacheWrite": 0 }, "contextWindow": 131072, - "maxTokens": 16384, - "supportsComputerUse": false + "maxTokens": 16384 }, "google/gemma-3-27b-it": { "id": "google/gemma-3-27b-it", @@ -31413,13 +31897,13 @@ "image" ], "cost": { - "input": 0.03, - "output": 0.11, - "cacheRead": 0.02, + "input": 0.08, + "output": 0.16, + "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 128000, - "maxTokens": 65536 + "contextWindow": 131072, + "maxTokens": 131072 }, "google/gemma-3-4b-it": { "id": "google/gemma-3-4b-it", @@ -31465,7 +31949,7 @@ }, "google/gemma-4-26b-a4b-it": { "id": "google/gemma-4-26b-a4b-it", - "name": "Gemma 4 26B A4B", + "name": "Gemma 4 26B A4B IT", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -31475,13 +31959,13 @@ "image" ], "cost": { - "input": 0.12, - "output": 0.4, + "input": 0.07, + "output": 0.34, "cacheRead": 0, "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 262144, + "maxTokens": 16384, "thinking": { "mode": "effort", "efforts": [ @@ -31495,7 +31979,7 @@ }, "google/gemma-4-31b-it": { "id": "google/gemma-4-31b-it", - "name": "Gemma 4 31B", + "name": "Gemma 4 31B IT", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -31505,13 +31989,13 @@ "image" ], "cost": { - "input": 0.14, - "output": 0.4, - "cacheRead": 0, + "input": 0.08, + "output": 0.35, + "cacheRead": 0.01, "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 131072, + "maxTokens": 262144, "thinking": { "mode": "effort", "efforts": [ @@ -31735,7 +32219,7 @@ }, "inclusionai/ling-2.6-flash": { "id": "inclusionai/ling-2.6-flash", - "name": "Ling-2.6 Flash", + "name": "Ling-2.6-flash", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -31744,9 +32228,9 @@ "text" ], "cost": { - "input": 0.08, - "output": 0.24, - "cacheRead": 0.016, + "input": 0.1, + "output": 0.3, + "cacheRead": 0.02, "cacheWrite": 0 }, "contextWindow": 262144, @@ -31773,6 +32257,35 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "inclusionai/ling-3.0-flash": { + "id": "inclusionai/ling-3.0-flash", + "name": "Ling-3.0-flash", + "api": "openai-completions", + "provider": "kilo", + "baseUrl": "https://api.kilo.ai/api/gateway", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.06, + "output": 0.18, + "cacheRead": 0.012, + "cacheWrite": 0 + }, + "contextWindow": 262144, + "maxTokens": 32768, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } + }, "inclusionai/ling-3.0-flash:free": { "id": "inclusionai/ling-3.0-flash:free", "name": "Ling-3.0-flash (free)", @@ -31791,7 +32304,37 @@ }, "contextWindow": 262144, "maxTokens": 32768, - "supportsComputerUse": false + "supportsComputerUse": false, + "supportsComputerUseConfig": false + }, + "inclusionai/ling-3.0-tiny:free": { + "id": "inclusionai/ling-3.0-tiny:free", + "name": "Ling 3.0 Tiny (free)", + "api": "openai-completions", + "provider": "kilo", + "baseUrl": "https://api.kilo.ai/api/gateway", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0, + "output": 0, + "cacheRead": 0, + "cacheWrite": 0 + }, + "contextWindow": 262144, + "maxTokens": 32768, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "inclusionai/ring-2.6-1t": { "id": "inclusionai/ring-2.6-1t", @@ -31804,9 +32347,9 @@ "text" ], "cost": { - "input": 0.075, - "output": 0.625, - "cacheRead": 0.015, + "input": 0.3, + "output": 2.5, + "cacheRead": 0.06, "cacheWrite": 0 }, "contextWindow": 262144, @@ -31887,22 +32430,23 @@ }, "kilo-auto/balanced": { "id": "kilo-auto/balanced", - "name": "Kilo Auto Balanced", + "name": "Auto Balanced", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0.6, - "output": 3, - "cacheRead": 0, - "cacheWrite": 0 + "input": 0.325, + "output": 1.95, + "cacheRead": 0.0325, + "cacheWrite": 0.40625 }, - "contextWindow": 204800, - "maxTokens": 131072, + "contextWindow": 1000000, + "maxTokens": 65536, "thinking": { "mode": "effort", "efforts": [ @@ -31920,23 +32464,33 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 0.325, + "output": 1.95, + "cacheRead": 0.0325, + "cacheWrite": 0.40625 }, - "contextWindow": null, - "maxTokens": null, - "supportsComputerUse": false + "contextWindow": 1000000, + "maxTokens": 65536, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "kilo-auto/free": { "id": "kilo-auto/free", - "name": "Kilo Auto Free", + "name": "Auto Free", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -31950,8 +32504,8 @@ "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 204800, - "maxTokens": 131072, + "contextWindow": 256000, + "maxTokens": 10000, "thinking": { "mode": "effort", "efforts": [ @@ -31965,7 +32519,7 @@ }, "kilo-auto/frontier": { "id": "kilo-auto/frontier", - "name": "Kilo Auto Frontier", + "name": "Auto Frontier", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -31977,8 +32531,8 @@ "cost": { "input": 5, "output": 25, - "cacheRead": 0, - "cacheWrite": 0 + "cacheRead": 0.5, + "cacheWrite": 6.25 }, "contextWindow": 1000000, "maxTokens": 128000, @@ -31995,7 +32549,7 @@ }, "kilo-auto/small": { "id": "kilo-auto/small", - "name": "Kilo Auto Small", + "name": "Auto Small", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -32007,11 +32561,11 @@ "cost": { "input": 0.05, "output": 0.4, - "cacheRead": 0, + "cacheRead": 0.005, "cacheWrite": 0 }, - "contextWindow": 400000, - "maxTokens": 128000, + "contextWindow": 262144, + "maxTokens": 32768, "thinking": { "mode": "effort", "efforts": [ @@ -32097,14 +32651,13 @@ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.15, + "output": 0.6, + "cacheRead": 0.03, "cacheWrite": 0 }, - "contextWindow": null, - "maxTokens": null, - "supportsComputerUse": false + "contextWindow": 256000, + "maxTokens": 80000 }, "kwaipilot/kat-coder-pro": { "id": "kwaipilot/kat-coder-pro", @@ -32157,14 +32710,13 @@ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.74, + "output": 2.96, + "cacheRead": 0.15, "cacheWrite": 0 }, - "contextWindow": null, - "maxTokens": null, - "supportsComputerUse": false + "contextWindow": 256000, + "maxTokens": 80000 }, "kwaipilot/kat-coder-pro-v2.5:free": { "id": "kwaipilot/kat-coder-pro-v2.5:free", @@ -32276,19 +32828,28 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.75, + "output": 3, + "cacheRead": 0.015, "cacheWrite": 0 }, - "contextWindow": null, - "maxTokens": null, - "supportsComputerUse": false + "contextWindow": 1048756, + "maxTokens": 262144, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "meituan/longcat-flash-chat": { "id": "meituan/longcat-flash-chat", @@ -32349,7 +32910,8 @@ "cacheWrite": 0 }, "contextWindow": 8192, - "maxTokens": 16384 + "maxTokens": 16384, + "supportsComputerUse": false }, "meta-llama/llama-3.1-405b": { "id": "meta-llama/llama-3.1-405b", @@ -32410,7 +32972,7 @@ "cacheWrite": 0 }, "contextWindow": 131072, - "maxTokens": 26215 + "maxTokens": 16384 }, "meta-llama/llama-3.1-8b-instruct": { "id": "meta-llama/llama-3.1-8b-instruct", @@ -32424,12 +32986,12 @@ ], "cost": { "input": 0.02, - "output": 0.05, + "output": 0.04, "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 16384, - "maxTokens": 16384 + "contextWindow": 131072, + "maxTokens": 131072 }, "meta-llama/llama-3.2-11b-vision-instruct": { "id": "meta-llama/llama-3.2-11b-vision-instruct", @@ -32494,7 +33056,7 @@ }, "meta-llama/llama-3.3-70b-instruct": { "id": "meta-llama/llama-3.3-70b-instruct", - "name": "Llama 3.3 70B Instruct", + "name": "Llama-3.3-70B-Instruct", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -32523,8 +33085,8 @@ "image" ], "cost": { - "input": 0.15, - "output": 0.6, + "input": 0.2, + "output": 0.696, "cacheRead": 0, "cacheWrite": 0 }, @@ -32543,7 +33105,7 @@ "image" ], "cost": { - "input": 0.08, + "input": 0.1, "output": 0.3, "cacheRead": 0, "cacheWrite": 0 @@ -32640,19 +33202,59 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 1.25, + "output": 4.25, + "cacheRead": 0.15, "cacheWrite": 0 }, "contextWindow": 1048576, "maxTokens": 1048576, - "supportsComputerUse": false + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } + }, + "meta/muse-spark-1.2": { + "id": "meta/muse-spark-1.2", + "name": "Muse Spark 1.2", + "api": "openai-completions", + "provider": "kilo", + "baseUrl": "https://api.kilo.ai/api/gateway", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 1.25, + "output": 4.25, + "cacheRead": 0.15, + "cacheWrite": 0 + }, + "contextWindow": 1048576, + "maxTokens": 1048576, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "microsoft/phi-4": { "id": "microsoft/phi-4", @@ -32766,7 +33368,7 @@ }, "minimax/minimax-m2": { "id": "minimax/minimax-m2", - "name": "MiniMax M2", + "name": "MiniMax-M2", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -32775,13 +33377,13 @@ "text" ], "cost": { - "input": 0.255, - "output": 1, - "cacheRead": 0.03, + "input": 0.3, + "output": 1.2, + "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 196608, - "maxTokens": 196608, + "contextWindow": 204800, + "maxTokens": 131072, "thinking": { "mode": "effort", "efforts": [ @@ -32814,7 +33416,7 @@ }, "minimax/minimax-m2.1": { "id": "minimax/minimax-m2.1", - "name": "MiniMax M2.1", + "name": "MiniMax-M2.1", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -32823,13 +33425,13 @@ "text" ], "cost": { - "input": 0.27, - "output": 0.95, + "input": 0.3, + "output": 1.2, "cacheRead": 0.03, "cacheWrite": 0 }, - "contextWindow": 196608, - "maxTokens": 39322, + "contextWindow": 204800, + "maxTokens": 131072, "thinking": { "mode": "effort", "efforts": [ @@ -32842,7 +33444,7 @@ }, "minimax/minimax-m2.5": { "id": "minimax/minimax-m2.5", - "name": "MiniMax M2.5", + "name": "MiniMax-M2.5", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -32851,9 +33453,9 @@ "text" ], "cost": { - "input": 0.25, + "input": 0.3, "output": 1.2, - "cacheRead": 0.029, + "cacheRead": 0.03, "cacheWrite": 0 }, "contextWindow": 196608, @@ -32891,7 +33493,7 @@ }, "minimax/minimax-m2.7": { "id": "minimax/minimax-m2.7", - "name": "MiniMax M2.7", + "name": "MiniMax-M2.7", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -32919,7 +33521,7 @@ }, "minimax/minimax-m3": { "id": "minimax/minimax-m3", - "name": "MiniMax M3", + "name": "MiniMax-M3", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -32934,7 +33536,7 @@ "cacheRead": 0.06, "cacheWrite": 0 }, - "contextWindow": 1048576, + "contextWindow": 524288, "maxTokens": 512000, "thinking": { "mode": "effort", @@ -32981,7 +33583,7 @@ "cost": { "input": 0.3, "output": 0.9, - "cacheRead": 0, + "cacheRead": 0.03, "cacheWrite": 0 }, "contextWindow": 256000, @@ -33004,7 +33606,8 @@ "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 65536 + "maxTokens": 65536, + "supportsComputerUse": false }, "mistralai/devstral-medium": { "id": "mistralai/devstral-medium", @@ -33023,7 +33626,8 @@ "cacheWrite": 0 }, "contextWindow": 131072, - "maxTokens": 26215 + "maxTokens": 26215, + "supportsComputerUse": false }, "mistralai/devstral-small": { "id": "mistralai/devstral-small", @@ -33042,7 +33646,8 @@ "cacheWrite": 0 }, "contextWindow": 131072, - "maxTokens": 26215 + "maxTokens": 26215, + "supportsComputerUse": false }, "mistralai/ministral-14b-2512": { "id": "mistralai/ministral-14b-2512", @@ -33058,7 +33663,7 @@ "cost": { "input": 0.2, "output": 0.2, - "cacheRead": 0, + "cacheRead": 0.02, "cacheWrite": 0 }, "contextWindow": 262144, @@ -33078,7 +33683,7 @@ "cost": { "input": 0.1, "output": 0.1, - "cacheRead": 0, + "cacheRead": 0.01, "cacheWrite": 0 }, "contextWindow": 131072, @@ -33098,7 +33703,7 @@ "cost": { "input": 0.15, "output": 0.15, - "cacheRead": 0, + "cacheRead": 0.015, "cacheWrite": 0 }, "contextWindow": 262144, @@ -33180,7 +33785,7 @@ "cost": { "input": 2, "output": 6, - "cacheRead": 0, + "cacheRead": 0.2, "cacheWrite": 0 }, "contextWindow": 128000, @@ -33199,7 +33804,7 @@ "cost": { "input": 2, "output": 6, - "cacheRead": 0, + "cacheRead": 0.2, "cacheWrite": 0 }, "contextWindow": 131072, @@ -33222,11 +33827,12 @@ "cacheWrite": 0 }, "contextWindow": 131072, - "maxTokens": 26215 + "maxTokens": 26215, + "supportsComputerUse": false }, "mistralai/mistral-large-2512": { "id": "mistralai/mistral-large-2512", - "name": "Mistral Large 3 2512", + "name": "Mistral Large 3", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -33238,7 +33844,7 @@ "cost": { "input": 0.5, "output": 1.5, - "cacheRead": 0, + "cacheRead": 0.05, "cacheWrite": 0 }, "contextWindow": 262144, @@ -33258,7 +33864,7 @@ "cost": { "input": 0.4, "output": 2, - "cacheRead": 0, + "cacheRead": 0.04, "cacheWrite": 0 }, "contextWindow": 131072, @@ -33308,7 +33914,7 @@ "cost": { "input": 0.4, "output": 2, - "cacheRead": 0, + "cacheRead": 0.04, "cacheWrite": 0 }, "contextWindow": 131072, @@ -33325,8 +33931,8 @@ "text" ], "cost": { - "input": 0.02, - "output": 0.04, + "input": 0.019, + "output": 0.03, "cacheRead": 0, "cacheWrite": 0 }, @@ -33346,7 +33952,7 @@ "cost": { "input": 0.2, "output": 0.6, - "cacheRead": 0, + "cacheRead": 0.02, "cacheWrite": 0 }, "contextWindow": 32768, @@ -33369,7 +33975,8 @@ "cacheWrite": 0 }, "contextWindow": 32768, - "maxTokens": 16384 + "maxTokens": 16384, + "supportsComputerUse": false }, "mistralai/mistral-small-2603": { "id": "mistralai/mistral-small-2603", @@ -33433,13 +34040,13 @@ "image" ], "cost": { - "input": 0.06, - "output": 0.18, - "cacheRead": 0.03, + "input": 0.09375, + "output": 0.25, + "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 131072, - "maxTokens": 131072 + "contextWindow": 256000, + "maxTokens": 16384 }, "mistralai/mistral-small-creative": { "id": "mistralai/mistral-small-creative", @@ -33475,7 +34082,7 @@ "cost": { "input": 2, "output": 6, - "cacheRead": 0, + "cacheRead": 0.2, "cacheWrite": 0 }, "contextWindow": 65536, @@ -33520,7 +34127,8 @@ "cacheWrite": 0 }, "contextWindow": 131072, - "maxTokens": 32768 + "maxTokens": 32768, + "supportsComputerUse": false }, "mistralai/voxtral-small-24b-2507": { "id": "mistralai/voxtral-small-24b-2507", @@ -33535,7 +34143,7 @@ "cost": { "input": 0.1, "output": 0.3, - "cacheRead": 0, + "cacheRead": 0.01, "cacheWrite": 0 }, "contextWindow": 32000, @@ -33552,13 +34160,13 @@ "text" ], "cost": { - "input": 0.55, - "output": 2.2, + "input": 0.57, + "output": 2.3, "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 131000, - "maxTokens": 26215 + "contextWindow": 131072, + "maxTokens": 100352 }, "moonshotai/kimi-k2-0905": { "id": "moonshotai/kimi-k2-0905", @@ -33571,13 +34179,13 @@ "text" ], "cost": { - "input": 0.4, - "output": 2, - "cacheRead": 0.15, + "input": 0.6, + "output": 2.5, + "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 131072, - "maxTokens": 26215 + "contextWindow": 262144, + "maxTokens": 100352 }, "moonshotai/kimi-k2-0905:exacto": { "id": "moonshotai/kimi-k2-0905:exacto", @@ -33611,13 +34219,13 @@ "text" ], "cost": { - "input": 0.47, - "output": 2, - "cacheRead": 0.2, + "input": 0.6, + "output": 2.5, + "cacheRead": 0.15, "cacheWrite": 0 }, - "contextWindow": 131072, - "maxTokens": 65535, + "contextWindow": 262144, + "maxTokens": 100352, "thinking": { "mode": "effort", "efforts": [ @@ -33642,13 +34250,13 @@ "image" ], "cost": { - "input": 0.45, - "output": 2.2, - "cacheRead": 0, + "input": 0.6, + "output": 3, + "cacheRead": 0.1, "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 65535, + "maxTokens": 262144, "thinking": { "mode": "effort", "efforts": [ @@ -33693,13 +34301,13 @@ "image" ], "cost": { - "input": 0.75, - "output": 3.5, - "cacheRead": 0.375, + "input": 0.8, + "output": 3.4, + "cacheRead": 0.16, "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 65535, + "maxTokens": 262144, "thinking": { "mode": "effort", "efforts": [ @@ -33744,9 +34352,9 @@ "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.95, + "output": 4, + "cacheRead": 0.19, "cacheWrite": 0 }, "contextWindow": 262144, @@ -33760,8 +34368,7 @@ "high", "xhigh" ] - }, - "supportsComputerUse": false + } }, "moonshotai/kimi-k3": { "id": "moonshotai/kimi-k3", @@ -33775,13 +34382,13 @@ "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 3, + "output": 15, + "cacheRead": 0.3, "cacheWrite": 0 }, "contextWindow": 1048576, - "maxTokens": 131072, + "maxTokens": 1048576, "thinking": { "mode": "effort", "efforts": [ @@ -33794,8 +34401,7 @@ "max": "max" }, "requiresEffort": true - }, - "supportsComputerUse": false + } }, "morph-warp-grep-v2": { "id": "morph-warp-grep-v2", @@ -33917,7 +34523,8 @@ "cacheWrite": 0 }, "contextWindow": 131072, - "maxTokens": 163840 + "maxTokens": 163840, + "supportsComputerUse": false }, "nex-agi/nex-n2-mini": { "id": "nex-agi/nex-n2-mini", @@ -33925,19 +34532,29 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.025, + "output": 0.1, + "cacheRead": 0.0025, "cacheWrite": 0 }, "contextWindow": 262144, "maxTokens": 262144, - "supportsComputerUse": false + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "nex-agi/nex-n2-pro": { "id": "nex-agi/nex-n2-pro", @@ -33945,19 +34562,29 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.25, + "output": 1, + "cacheRead": 0.025, "cacheWrite": 0 }, "contextWindow": 262144, "maxTokens": 262144, - "supportsComputerUse": false + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "nex-agi/nex-n2-pro:free": { "id": "nex-agi/nex-n2-pro:free", @@ -34135,7 +34762,7 @@ }, "nvidia/llama-3.3-nemotron-super-49b-v1.5": { "id": "nvidia/llama-3.3-nemotron-super-49b-v1.5", - "name": "Llama 3.3 Nemotron Super 49B V1.5", + "name": "Llama 3.3 Nemotron Super 49B v1.5", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -34160,7 +34787,8 @@ "high", "xhigh" ] - } + }, + "supportsComputerUse": false }, "nvidia/nemotron-3-nano-30b-a3b": { "id": "nvidia/nemotron-3-nano-30b-a3b", @@ -34175,11 +34803,11 @@ "cost": { "input": 0.05, "output": 0.2, - "cacheRead": 0, + "cacheRead": 0.03, "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 52429, + "maxTokens": 262144, "thinking": { "mode": "effort", "efforts": [ @@ -34224,7 +34852,7 @@ }, "nvidia/nemotron-3-super-120b-a12b": { "id": "nvidia/nemotron-3-super-120b-a12b", - "name": "Nemotron 3 Super", + "name": "Nemotron 3 Super 120B A12B", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -34233,13 +34861,13 @@ "text" ], "cost": { - "input": 0.1, - "output": 0.5, - "cacheRead": 0.1, + "input": 0.085, + "output": 0.4, + "cacheRead": 0, "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 262144, + "maxTokens": 16384, "thinking": { "mode": "effort", "efforts": [ @@ -34290,6 +34918,35 @@ "input": [ "text" ], + "cost": { + "input": 0.5, + "output": 2.2, + "cacheRead": 0.1, + "cacheWrite": 0 + }, + "contextWindow": 512288, + "maxTokens": 512288, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } + }, + "nvidia/nemotron-3-ultra-550b-a55b:free": { + "id": "nvidia/nemotron-3-ultra-550b-a55b:free", + "name": "Nemotron 3 Ultra (free)", + "api": "openai-completions", + "provider": "kilo", + "baseUrl": "https://api.kilo.ai/api/gateway", + "reasoning": true, + "input": [ + "text" + ], "cost": { "input": 0, "output": 0, @@ -34307,28 +34964,7 @@ "high", "xhigh" ] - }, - "supportsComputerUse": false - }, - "nvidia/nemotron-3-ultra-550b-a55b:free": { - "id": "nvidia/nemotron-3-ultra-550b-a55b:free", - "name": "Nemotron 3 Ultra (free)", - "api": "openai-completions", - "provider": "kilo", - "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, - "input": [ - "text" - ], - "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 - }, - "contextWindow": null, - "maxTokens": null, - "supportsComputerUse": false + } }, "nvidia/nemotron-3.5-content-safety:free": { "id": "nvidia/nemotron-3.5-content-safety:free", @@ -34409,11 +35045,12 @@ "high", "xhigh" ] - } + }, + "supportsComputerUse": false }, "openai/gpt-3.5-turbo": { "id": "openai/gpt-3.5-turbo", - "name": "GPT-3.5 Turbo", + "name": "GPT-3.5-turbo", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -34524,7 +35161,8 @@ "cacheWrite": 0 }, "contextWindow": 8191, - "maxTokens": 4096 + "maxTokens": 4096, + "supportsComputerUse": false }, "openai/gpt-4-1106-preview": { "id": "openai/gpt-4-1106-preview", @@ -34543,7 +35181,8 @@ "cacheWrite": 0 }, "contextWindow": 128000, - "maxTokens": 4096 + "maxTokens": 4096, + "supportsComputerUse": false }, "openai/gpt-4-turbo": { "id": "openai/gpt-4-turbo", @@ -34606,7 +35245,7 @@ }, "openai/gpt-4.1-mini": { "id": "openai/gpt-4.1-mini", - "name": "GPT-4.1 Mini", + "name": "GPT-4.1 mini", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -34626,7 +35265,7 @@ }, "openai/gpt-4.1-nano": { "id": "openai/gpt-4.1-nano", - "name": "GPT-4.1 Nano", + "name": "GPT-4.1 nano", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -34741,11 +35380,12 @@ "cacheWrite": 0 }, "contextWindow": 128000, - "maxTokens": 16384 + "maxTokens": 16384, + "supportsComputerUse": false }, "openai/gpt-4o-mini": { "id": "openai/gpt-4o-mini", - "name": "GPT-4o-mini", + "name": "GPT-4o mini", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -34777,7 +35417,7 @@ "cost": { "input": 0.15, "output": 0.6, - "cacheRead": 0, + "cacheRead": 0.075, "cacheWrite": 0 }, "contextWindow": 128000, @@ -34923,7 +35563,8 @@ "medium", "high" ] - } + }, + "supportsComputerUse": false }, "openai/gpt-5-image": { "id": "openai/gpt-5-image", @@ -34952,7 +35593,8 @@ "medium", "high" ] - } + }, + "supportsComputerUse": false }, "openai/gpt-5-image-mini": { "id": "openai/gpt-5-image-mini", @@ -34981,7 +35623,8 @@ "medium", "high" ] - } + }, + "supportsComputerUse": false }, "openai/gpt-5-mini": { "id": "openai/gpt-5-mini", @@ -35117,11 +35760,12 @@ "cacheWrite": 0 }, "contextWindow": 128000, - "maxTokens": 16384 + "maxTokens": 16384, + "supportsComputerUse": false }, "openai/gpt-5.1-codex": { "id": "openai/gpt-5.1-codex", - "name": "GPT-5.1-Codex", + "name": "GPT-5.1 Codex", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -35133,7 +35777,7 @@ "cost": { "input": 1.25, "output": 10, - "cacheRead": 0.125, + "cacheRead": 0.13, "cacheWrite": 0 }, "contextWindow": 272000, @@ -35150,7 +35794,7 @@ }, "openai/gpt-5.1-codex-max": { "id": "openai/gpt-5.1-codex-max", - "name": "GPT-5.1-Codex-Max", + "name": "GPT-5.1 Codex Max", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -35179,7 +35823,7 @@ }, "openai/gpt-5.1-codex-mini": { "id": "openai/gpt-5.1-codex-mini", - "name": "GPT-5.1-Codex-Mini", + "name": "GPT-5.1 Codex mini", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -35191,11 +35835,11 @@ "cost": { "input": 0.25, "output": 2, - "cacheRead": 0.025, + "cacheRead": 0.03, "cacheWrite": 0 }, "contextWindow": 272000, - "maxTokens": 100000, + "maxTokens": 128000, "thinking": { "mode": "effort", "efforts": [ @@ -35255,7 +35899,7 @@ }, "openai/gpt-5.2-codex": { "id": "openai/gpt-5.2-codex", - "name": "GPT-5.2-Codex", + "name": "GPT-5.2 Codex", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -35325,7 +35969,7 @@ "cost": { "input": 1.75, "output": 14, - "cacheRead": 0, + "cacheRead": 0.175, "cacheWrite": 0 }, "contextWindow": 128000, @@ -35333,7 +35977,7 @@ }, "openai/gpt-5.3-codex": { "id": "openai/gpt-5.3-codex", - "name": "GPT-5.3-Codex", + "name": "GPT-5.3 Codex", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -35345,7 +35989,7 @@ "cost": { "input": 1.75, "output": 14, - "cacheRead": 0, + "cacheRead": 0.175, "cacheWrite": 0 }, "contextWindow": 272000, @@ -35374,7 +36018,7 @@ "cost": { "input": 2.5, "output": 15, - "cacheRead": 0, + "cacheRead": 0.25, "cacheWrite": 0 }, "contextWindow": 1050000, @@ -35411,7 +36055,7 @@ }, "openai/gpt-5.4-mini": { "id": "openai/gpt-5.4-mini", - "name": "GPT-5.4 Mini", + "name": "GPT-5.4 mini", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -35440,7 +36084,7 @@ }, "openai/gpt-5.4-nano": { "id": "openai/gpt-5.4-nano", - "name": "GPT-5.4 Nano", + "name": "GPT-5.4 nano", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -35568,12 +36212,12 @@ "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 0.2, + "output": 1.2, + "cacheRead": 0.02, + "cacheWrite": 0.25 }, - "contextWindow": 372000, + "contextWindow": 1050000, "maxTokens": 128000, "thinking": { "mode": "effort", @@ -35584,28 +36228,37 @@ "xhigh", "max" ] - }, - "supportsComputerUse": false + } }, "openai/gpt-5.6-luna-pro": { "id": "openai/gpt-5.6-luna-pro", - "name": "GPT-5.6 Luna Pro", + "name": "GPT-5.6 Luna", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 0.2, + "output": 1.2, + "cacheRead": 0.02, + "cacheWrite": 0.25 }, "contextWindow": 1050000, "maxTokens": 128000, - "supportsComputerUse": false + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "high", + "xhigh", + "max" + ] + } }, "openai/gpt-5.6-sol": { "id": "openai/gpt-5.6-sol", @@ -35619,12 +36272,12 @@ "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 5, + "output": 30, + "cacheRead": 0.5, + "cacheWrite": 6.25 }, - "contextWindow": 372000, + "contextWindow": 1050000, "maxTokens": 128000, "thinking": { "mode": "effort", @@ -35635,28 +36288,37 @@ "xhigh", "max" ] - }, - "supportsComputerUse": false + } }, "openai/gpt-5.6-sol-pro": { "id": "openai/gpt-5.6-sol-pro", - "name": "GPT-5.6 Sol Pro", + "name": "GPT-5.6 Sol", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 5, + "output": 30, + "cacheRead": 0.5, + "cacheWrite": 6.25 }, "contextWindow": 1050000, "maxTokens": 128000, - "supportsComputerUse": false + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "high", + "xhigh", + "max" + ] + } }, "openai/gpt-5.6-terra": { "id": "openai/gpt-5.6-terra", @@ -35670,12 +36332,12 @@ "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 2, + "output": 12, + "cacheRead": 0.2, + "cacheWrite": 2.5 }, - "contextWindow": 372000, + "contextWindow": 1050000, "maxTokens": 128000, "thinking": { "mode": "effort", @@ -35686,28 +36348,37 @@ "xhigh", "max" ] - }, - "supportsComputerUse": false + } }, "openai/gpt-5.6-terra-pro": { "id": "openai/gpt-5.6-terra-pro", - "name": "GPT-5.6 Terra Pro", + "name": "GPT-5.6 Terra", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 2, + "output": 12, + "cacheRead": 0.2, + "cacheWrite": 2.5 }, "contextWindow": 1050000, "maxTokens": 128000, - "supportsComputerUse": false + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "high", + "xhigh", + "max" + ] + } }, "openai/gpt-audio": { "id": "openai/gpt-audio", @@ -35720,14 +36391,13 @@ "text" ], "cost": { - "input": 0, - "output": 0, + "input": 2.5, + "output": 10, "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": null, - "maxTokens": null, - "supportsComputerUse": false + "contextWindow": 128000, + "maxTokens": 16384 }, "openai/gpt-audio-mini": { "id": "openai/gpt-audio-mini", @@ -35740,14 +36410,13 @@ "text" ], "cost": { - "input": 0, - "output": 0, + "input": 0.6, + "output": 2.4, "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": null, - "maxTokens": null, - "supportsComputerUse": false + "contextWindow": 128000, + "maxTokens": 16384 }, "openai/gpt-chat-latest": { "id": "openai/gpt-chat-latest", @@ -35755,7 +36424,7 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": true, + "reasoning": false, "input": [ "text", "image" @@ -35767,21 +36436,11 @@ "cacheWrite": 0 }, "contextWindow": 400000, - "maxTokens": 128000, - "thinking": { - "mode": "effort", - "efforts": [ - "minimal", - "low", - "medium", - "high", - "xhigh" - ] - } + "maxTokens": 128000 }, "openai/gpt-oss-120b": { "id": "openai/gpt-oss-120b", - "name": "gpt-oss-120b", + "name": "GPT OSS 120B", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -35790,13 +36449,13 @@ "text" ], "cost": { - "input": 0.039, - "output": 0.19, - "cacheRead": 0, + "input": 0.03, + "output": 0.17, + "cacheRead": 0.03, "cacheWrite": 0 }, "contextWindow": 131072, - "maxTokens": 26215, + "maxTokens": 131072, "thinking": { "mode": "effort", "efforts": [ @@ -35829,7 +36488,7 @@ }, "openai/gpt-oss-20b": { "id": "openai/gpt-oss-20b", - "name": "gpt-oss-20b", + "name": "GPT OSS 20B", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -35839,12 +36498,12 @@ ], "cost": { "input": 0.03, - "output": 0.14, - "cacheRead": 0, + "output": 0.13, + "cacheRead": 0.03, "cacheWrite": 0 }, "contextWindow": 131072, - "maxTokens": 26215, + "maxTokens": 131072, "thinking": { "mode": "effort", "efforts": [ @@ -35867,7 +36526,7 @@ "cost": { "input": 0.075, "output": 0.3, - "cacheRead": 0.037, + "cacheRead": 0.0375, "cacheWrite": 0 }, "contextWindow": 131072, @@ -35887,7 +36546,7 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ "text", "image" @@ -35899,7 +36558,18 @@ "cacheWrite": 0 }, "contextWindow": 200000, - "maxTokens": 100000 + "maxTokens": 100000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ], + "requiresEffort": true + } }, "openai/o1-pro": { "id": "openai/o1-pro", @@ -35981,15 +36651,16 @@ "xhigh" ], "requiresEffort": true - } + }, + "supportsComputerUse": false }, "openai/o3-mini": { "id": "openai/o3-mini", - "name": "o3 Mini", + "name": "o3-mini", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ "text" ], @@ -36000,7 +36671,18 @@ "cacheWrite": 0 }, "contextWindow": 200000, - "maxTokens": 100000 + "maxTokens": 100000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ], + "requiresEffort": true + } }, "openai/o3-mini-high": { "id": "openai/o3-mini-high", @@ -36008,7 +36690,7 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ "text" ], @@ -36019,11 +36701,22 @@ "cacheWrite": 0 }, "contextWindow": 200000, - "maxTokens": 100000 + "maxTokens": 100000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ], + "requiresEffort": true + } }, "openai/o3-pro": { "id": "openai/o3-pro", - "name": "o3 Pro", + "name": "o3-pro", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -36054,7 +36747,7 @@ }, "openai/o4-mini": { "id": "openai/o4-mini", - "name": "o4 Mini", + "name": "o4-mini", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -36112,7 +36805,8 @@ "xhigh" ], "requiresEffort": true - } + }, + "supportsComputerUse": false }, "openai/o4-mini-high": { "id": "openai/o4-mini-high", @@ -36128,7 +36822,7 @@ "cost": { "input": 1.1, "output": 4.4, - "cacheRead": 0, + "cacheRead": 0.275, "cacheWrite": 0 }, "contextWindow": 200000, @@ -36259,7 +36953,7 @@ }, "openrouter/free": { "id": "openrouter/free", - "name": "Free Models Router", + "name": "OpenRouter Free Models Router", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -36376,7 +37070,8 @@ "high", "xhigh" ] - } + }, + "supportsComputerUse": false }, "openrouter/pareto-code": { "id": "openrouter/pareto-code", @@ -36574,7 +37269,8 @@ "high", "xhigh" ] - } + }, + "supportsComputerUse": false }, "poolside/laguna-s-2.1": { "id": "poolside/laguna-s-2.1", @@ -36582,19 +37278,28 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.1, + "output": 0.2, + "cacheRead": 0.01, "cacheWrite": 0 }, "contextWindow": 1048576, "maxTokens": 131072, - "supportsComputerUse": false + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "poolside/laguna-s-2.1:free": { "id": "poolside/laguna-s-2.1:free", @@ -36602,26 +37307,6 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, - "input": [ - "text" - ], - "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 - }, - "contextWindow": 1048576, - "maxTokens": 131072, - "supportsComputerUse": false - }, - "poolside/laguna-xs-2.1": { - "id": "poolside/laguna-xs-2.1", - "name": "Laguna XS 2.1", - "api": "openai-completions", - "provider": "kilo", - "baseUrl": "https://api.kilo.ai/api/gateway", "reasoning": true, "input": [ "text" @@ -36634,7 +37319,35 @@ }, "contextWindow": 262144, "maxTokens": 32768, - "supportsComputerUse": false, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } + }, + "poolside/laguna-xs-2.1": { + "id": "poolside/laguna-xs-2.1", + "name": "Laguna XS 2.1", + "api": "openai-completions", + "provider": "kilo", + "baseUrl": "https://api.kilo.ai/api/gateway", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.1, + "output": 0.2, + "cacheRead": 0.05, + "cacheWrite": 0 + }, + "contextWindow": 262144, + "maxTokens": 32768, "thinking": { "mode": "effort", "efforts": [ @@ -36652,7 +37365,7 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ "text" ], @@ -36664,7 +37377,16 @@ }, "contextWindow": 262144, "maxTokens": 32768, - "supportsComputerUse": false + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "poolside/laguna-xs.2": { "id": "poolside/laguna-xs.2", @@ -36724,7 +37446,8 @@ "high", "xhigh" ] - } + }, + "supportsComputerUse": false }, "prime-intellect/intellect-3": { "id": "prime-intellect/intellect-3", @@ -36753,7 +37476,8 @@ "high", "xhigh" ] - } + }, + "supportsComputerUse": false }, "qwen/qwen-2.5-72b-instruct": { "id": "qwen/qwen-2.5-72b-instruct", @@ -36766,8 +37490,8 @@ "text" ], "cost": { - "input": 0.12, - "output": 0.39, + "input": 0.36, + "output": 0.4, "cacheRead": 0, "cacheWrite": 0 }, @@ -36785,13 +37509,13 @@ "text" ], "cost": { - "input": 0.04, - "output": 0.1, + "input": 0.1, + "output": 0.2, "cacheRead": 0, "cacheWrite": 0 }, "contextWindow": 32768, - "maxTokens": 6554 + "maxTokens": 32768 }, "qwen/qwen-2.5-coder-32b-instruct": { "id": "qwen/qwen-2.5-coder-32b-instruct", @@ -36857,7 +37581,7 @@ }, "qwen/qwen-plus": { "id": "qwen/qwen-plus", - "name": "Qwen-Plus", + "name": "Qwen Plus", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -36866,10 +37590,10 @@ "text" ], "cost": { - "input": 0.4, - "output": 1.2, - "cacheRead": 0.08, - "cacheWrite": 0 + "input": 0.26, + "output": 0.78, + "cacheRead": 0.052, + "cacheWrite": 0.325 }, "contextWindow": 1000000, "maxTokens": 32768 @@ -36904,10 +37628,10 @@ "text" ], "cost": { - "input": 0.26, - "output": 0.78, + "input": 0.4, + "output": 1.2, "cacheRead": 0, - "cacheWrite": 0 + "cacheWrite": 0.5 }, "contextWindow": 1000000, "maxTokens": 32768, @@ -37057,13 +37781,13 @@ "text" ], "cost": { - "input": 0.06, - "output": 0.24, - "cacheRead": 0.025, + "input": 0.2275, + "output": 0.91, + "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 40960, - "maxTokens": 40960, + "contextWindow": 131072, + "maxTokens": 8192, "thinking": { "mode": "effort", "efforts": [ @@ -37076,7 +37800,7 @@ }, "qwen/qwen3-235b-a22b": { "id": "qwen/qwen3-235b-a22b", - "name": "Qwen3 235B A22B", + "name": "Qwen3 235B-A22B", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -37087,7 +37811,7 @@ "cost": { "input": 0.455, "output": 1.82, - "cacheRead": 0.15, + "cacheRead": 0, "cacheWrite": 0 }, "contextWindow": 131072, @@ -37113,13 +37837,13 @@ "text" ], "cost": { - "input": 0.071, - "output": 0.1, + "input": 0.1495, + "output": 0.598, "cacheRead": 0, "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 52429 + "maxTokens": 16384 }, "qwen/qwen3-235b-a22b-thinking-2507": { "id": "qwen/qwen3-235b-a22b-thinking-2507", @@ -37132,12 +37856,12 @@ "text" ], "cost": { - "input": 0.11, - "output": 0.6, + "input": 0.23, + "output": 2.3, "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 262144, + "contextWindow": 131072, "maxTokens": 262144, "thinking": { "mode": "effort", @@ -37161,13 +37885,13 @@ "text" ], "cost": { - "input": 0.08, - "output": 0.28, - "cacheRead": 0.03, + "input": 0.13, + "output": 0.52, + "cacheRead": 0, "cacheWrite": 0 }, "contextWindow": 40960, - "maxTokens": 40960, + "maxTokens": 16384, "thinking": { "mode": "effort", "efforts": [ @@ -37189,13 +37913,13 @@ "text" ], "cost": { - "input": 0.09, - "output": 0.3, - "cacheRead": 0.04, + "input": 0.13, + "output": 0.52, + "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 262144, - "maxTokens": 262144 + "contextWindow": 128000, + "maxTokens": 32000 }, "qwen/qwen3-30b-a3b-thinking-2507": { "id": "qwen/qwen3-30b-a3b-thinking-2507", @@ -37208,13 +37932,13 @@ "text" ], "cost": { - "input": 0.051, - "output": 0.34, + "input": 0.2, + "output": 2.4, "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 32768, - "maxTokens": 6554, + "contextWindow": 81920, + "maxTokens": 32768, "thinking": { "mode": "effort", "efforts": [ @@ -37237,13 +37961,13 @@ "text" ], "cost": { - "input": 0.08, - "output": 0.24, - "cacheRead": 0.04, + "input": 0.104, + "output": 0.416, + "cacheRead": 0, "cacheWrite": 0 }, "contextWindow": 40960, - "maxTokens": 40960, + "maxTokens": 16384, "thinking": { "mode": "effort", "efforts": [ @@ -37265,12 +37989,12 @@ "text" ], "cost": { - "input": 0.05, - "output": 0.4, - "cacheRead": 0.05, + "input": 0.117, + "output": 0.455, + "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 40960, + "contextWindow": 131072, "maxTokens": 8192, "thinking": { "mode": "effort", @@ -37293,17 +38017,17 @@ "text" ], "cost": { - "input": 0.22, - "output": 1, - "cacheRead": 0.022, + "input": 0.975, + "output": 4.875, + "cacheRead": 0, "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 52429 + "maxTokens": 65536 }, "qwen/qwen3-coder-30b-a3b-instruct": { "id": "qwen/qwen3-coder-30b-a3b-instruct", - "name": "Qwen3 Coder 30B A3B Instruct", + "name": "Qwen3-Coder 30B-A3B Instruct", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -37312,8 +38036,8 @@ "text" ], "cost": { - "input": 0.07, - "output": 0.27, + "input": 0.2925, + "output": 1.4625, "cacheRead": 0, "cacheWrite": 0 }, @@ -37333,8 +38057,8 @@ "cost": { "input": 0.195, "output": 0.975, - "cacheRead": 0.06, - "cacheWrite": 0 + "cacheRead": 0.039, + "cacheWrite": 0.24375 }, "contextWindow": 1000000, "maxTokens": 65536 @@ -37350,13 +38074,13 @@ "text" ], "cost": { - "input": 0.12, - "output": 0.75, - "cacheRead": 0.035, + "input": 0.3, + "output": 1.5, + "cacheRead": 0, "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 65536 + "maxTokens": 262144 }, "qwen/qwen3-coder-plus": { "id": "qwen/qwen3-coder-plus", @@ -37371,8 +38095,8 @@ "cost": { "input": 0.65, "output": 3.25, - "cacheRead": 0.2, - "cacheWrite": 0 + "cacheRead": 0.13, + "cacheWrite": 0.8125 }, "contextWindow": 1000000, "maxTokens": 65536 @@ -37409,13 +38133,13 @@ "text" ], "cost": { - "input": 1.2, - "output": 6, - "cacheRead": 0.24, - "cacheWrite": 0 + "input": 0.78, + "output": 3.9, + "cacheRead": 0.156, + "cacheWrite": 0.975 }, "contextWindow": 262144, - "maxTokens": 32768 + "maxTokens": 65536 }, "qwen/qwen3-max-thinking": { "id": "qwen/qwen3-max-thinking", @@ -37434,7 +38158,7 @@ "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 32768, + "maxTokens": 65536, "thinking": { "mode": "effort", "efforts": [ @@ -37448,7 +38172,7 @@ }, "qwen/qwen3-next-80b-a3b-instruct": { "id": "qwen/qwen3-next-80b-a3b-instruct", - "name": "Qwen3 Next 80B A3B Instruct", + "name": "Qwen3-Next 80B-A3B Instruct", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -37456,33 +38180,33 @@ "input": [ "text" ], - "cost": { - "input": 0.09, - "output": 1.1, - "cacheRead": 0, - "cacheWrite": 0 - }, - "contextWindow": 131072, - "maxTokens": 52429 - }, - "qwen/qwen3-next-80b-a3b-thinking": { - "id": "qwen/qwen3-next-80b-a3b-thinking", - "name": "Qwen3 Next 80B A3B Thinking", - "api": "openai-completions", - "provider": "kilo", - "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": true, - "input": [ - "text" - ], "cost": { "input": 0.0975, "output": 0.78, "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 131072, - "maxTokens": 32768, + "contextWindow": 262144, + "maxTokens": 16384 + }, + "qwen/qwen3-next-80b-a3b-thinking": { + "id": "qwen/qwen3-next-80b-a3b-thinking", + "name": "Qwen3-Next 80B-A3B (Thinking)", + "api": "openai-completions", + "provider": "kilo", + "baseUrl": "https://api.kilo.ai/api/gateway", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.15, + "output": 1.2, + "cacheRead": 0, + "cacheWrite": 0 + }, + "contextWindow": 262144, + "maxTokens": 262144, "thinking": { "mode": "effort", "efforts": [ @@ -37506,13 +38230,13 @@ "image" ], "cost": { - "input": 0.2, - "output": 0.88, - "cacheRead": 0.11, + "input": 0.26, + "output": 1.04, + "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 262144, - "maxTokens": 52429 + "contextWindow": 131072, + "maxTokens": 32768 }, "qwen/qwen3-vl-235b-a22b-thinking": { "id": "qwen/qwen3-vl-235b-a22b-thinking", @@ -37526,8 +38250,8 @@ "image" ], "cost": { - "input": 0.26, - "output": 2.6, + "input": 0.4, + "output": 4, "cacheRead": 0, "cacheWrite": 0 }, @@ -37561,8 +38285,8 @@ "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 131072, - "maxTokens": 32768 + "contextWindow": 262144, + "maxTokens": 16384 }, "qwen/qwen3-vl-30b-a3b-thinking": { "id": "qwen/qwen3-vl-30b-a3b-thinking", @@ -37576,8 +38300,8 @@ "image" ], "cost": { - "input": 0.13, - "output": 1.56, + "input": 0.2, + "output": 2.4, "cacheRead": 0, "cacheWrite": 0 }, @@ -37626,8 +38350,8 @@ "image" ], "cost": { - "input": 0.08, - "output": 0.5, + "input": 0.117, + "output": 0.455, "cacheRead": 0, "cacheWrite": 0 }, @@ -37646,8 +38370,8 @@ "image" ], "cost": { - "input": 0.117, - "output": 1.365, + "input": 0.18, + "output": 2.1, "cacheRead": 0, "cacheWrite": 0 }, @@ -37666,7 +38390,7 @@ }, "qwen/qwen3.5-122b-a10b": { "id": "qwen/qwen3.5-122b-a10b", - "name": "Qwen3.5-122B-A10B", + "name": "Qwen3.5 122B-A10B", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -37682,7 +38406,7 @@ "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 65536, + "maxTokens": 81920, "thinking": { "mode": "effort", "efforts": [ @@ -37695,7 +38419,7 @@ }, "qwen/qwen3.5-27b": { "id": "qwen/qwen3.5-27b", - "name": "Qwen3.5-27B", + "name": "Qwen3.5 27B", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -37724,7 +38448,7 @@ }, "qwen/qwen3.5-35b-a3b": { "id": "qwen/qwen3.5-35b-a3b", - "name": "Qwen3.5-35B-A3B", + "name": "Qwen3.5 35B-A3B", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -37740,7 +38464,7 @@ "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 65536, + "maxTokens": 262144, "thinking": { "mode": "effort", "efforts": [ @@ -37753,7 +38477,7 @@ }, "qwen/qwen3.5-397b-a17b": { "id": "qwen/qwen3.5-397b-a17b", - "name": "Qwen3.5 397B A17B", + "name": "Qwen3.5 397B-A17B", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -37782,7 +38506,7 @@ }, "qwen/qwen3.5-9b": { "id": "qwen/qwen3.5-9b", - "name": "Qwen3.5-9B", + "name": "Qwen3.5 9B", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -37792,13 +38516,13 @@ "image" ], "cost": { - "input": 0.05, + "input": 0.1, "output": 0.15, "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 256000, - "maxTokens": 32768, + "contextWindow": 262144, + "maxTokens": 262144, "thinking": { "mode": "effort", "efforts": [ @@ -37821,8 +38545,8 @@ "image" ], "cost": { - "input": 0.1, - "output": 0.4, + "input": 0.065, + "output": 0.26, "cacheRead": 0, "cacheWrite": 0 }, @@ -37879,10 +38603,10 @@ "image" ], "cost": { - "input": 0.4, - "output": 2.4, + "input": 0.3, + "output": 1.8, "cacheRead": 0, - "cacheWrite": 0 + "cacheWrite": 0.375 }, "contextWindow": 1000000, "maxTokens": 65536, @@ -37908,13 +38632,13 @@ "image" ], "cost": { - "input": 0.325, - "output": 3.25, + "input": 0.45, + "output": 2.7, "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 256000, - "maxTokens": 65536, + "contextWindow": 262144, + "maxTokens": 262144, "thinking": { "mode": "effort", "efforts": [ @@ -37927,23 +38651,32 @@ }, "qwen/qwen3.6-35b-a3b": { "id": "qwen/qwen3.6-35b-a3b", - "name": "Qwen3.6 35B A3B", + "name": "Qwen3.6 35B-A3B", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.14, + "output": 1, + "cacheRead": 0.05, "cacheWrite": 0 }, - "contextWindow": 256000, - "maxTokens": 65536, - "supportsComputerUse": false + "contextWindow": 262144, + "maxTokens": 262144, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + } }, "qwen/qwen3.6-flash": { "id": "qwen/qwen3.6-flash", @@ -37957,10 +38690,10 @@ "image" ], "cost": { - "input": 0.25, - "output": 1.5, + "input": 0.1875, + "output": 1.125, "cacheRead": 0, - "cacheWrite": 0.3125 + "cacheWrite": 0.234375 }, "contextWindow": 1000000, "maxTokens": 65536, @@ -37985,10 +38718,10 @@ "text" ], "cost": { - "input": 1.04, - "output": 6.24, + "input": 1.027, + "output": 6.162, "cacheRead": 0, - "cacheWrite": 1.3 + "cacheWrite": 1.28375 }, "contextWindow": 262144, "maxTokens": 65536, @@ -38016,7 +38749,7 @@ "cost": { "input": 0.325, "output": 1.95, - "cacheRead": 0.0325, + "cacheRead": 0, "cacheWrite": 0.40625 }, "contextWindow": 1000000, @@ -38079,19 +38812,28 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 0.03, + "output": 0.13, + "cacheRead": 0.006, + "cacheWrite": 0.038 }, "contextWindow": 1000000, "maxTokens": 65536, - "supportsComputerUse": false + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + } }, "qwen/qwen3.7-max": { "id": "qwen/qwen3.7-max", @@ -38104,13 +38846,13 @@ "text" ], "cost": { - "input": 1.625, - "output": 4.875, - "cacheRead": 0.1625, - "cacheWrite": 2.03125 + "input": 1.25, + "output": 3.75, + "cacheRead": 0.125, + "cacheWrite": 1.5625 }, "contextWindow": 1000000, - "maxTokens": 65536, + "maxTokens": 131072, "thinking": { "mode": "effort", "efforts": [ @@ -38133,13 +38875,13 @@ "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 0.32, + "output": 1.28, + "cacheRead": 0.032, + "cacheWrite": 0.4 }, "contextWindow": 1000000, - "maxTokens": 64000, + "maxTokens": 131072, "thinking": { "mode": "effort", "efforts": [ @@ -38148,8 +38890,7 @@ "medium", "high" ] - }, - "supportsComputerUse": false + } }, "qwen/qwen3.7-plus:free": { "id": "qwen/qwen3.7-plus:free", @@ -38172,6 +38913,35 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "qwen/qwen3.8-max": { + "id": "qwen/qwen3.8-max", + "name": "Qwen3.8 Max", + "api": "openai-completions", + "provider": "kilo", + "baseUrl": "https://api.kilo.ai/api/gateway", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 2, + "output": 6, + "cacheRead": 0.25, + "cacheWrite": 2.5 + }, + "contextWindow": 1000000, + "maxTokens": 131072, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + } + }, "qwen/qwq-32b": { "id": "qwen/qwq-32b", "name": "QwQ 32B", @@ -38320,20 +39090,29 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 5, + "output": 30, + "cacheRead": 0.5, "cacheWrite": 0 }, "contextWindow": 1000000, - "maxTokens": null, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "sao10k/l3-euryale-70b": { "id": "sao10k/l3-euryale-70b", @@ -38352,7 +39131,8 @@ "cacheWrite": 0 }, "contextWindow": 8192, - "maxTokens": 8192 + "maxTokens": 8192, + "supportsComputerUse": false }, "sao10k/l3-lunaris-8b": { "id": "sao10k/l3-lunaris-8b", @@ -38500,19 +39280,29 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 4, + "output": 20, + "cacheRead": 0.4, + "cacheWrite": 5 }, "contextWindow": 1000000, "maxTokens": 128000, - "supportsComputerUse": false + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "stealth/claude-sonnet-4.6": { "id": "stealth/claude-sonnet-4.6", @@ -38550,19 +39340,29 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 4, + "output": 24, + "cacheRead": 0.4, + "cacheWrite": 5 }, "contextWindow": 1050000, "maxTokens": 128000, - "supportsComputerUse": false + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "high", + "xhigh", + "max" + ] + } }, "stealth/qwen3.6-plus": { "id": "stealth/qwen3.6-plus", @@ -38570,19 +39370,28 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 0.25, + "output": 1.5, + "cacheRead": 0.025, + "cacheWrite": 0.3125 }, "contextWindow": 1000000, "maxTokens": 65536, - "supportsComputerUse": false + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + } }, "stepfun/step-3.5-flash": { "id": "stepfun/step-3.5-flash", @@ -38597,11 +39406,11 @@ "cost": { "input": 0.1, "output": 0.3, - "cacheRead": 0.02, + "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 256000, - "maxTokens": 256000, + "contextWindow": 262144, + "maxTokens": 65536, "thinking": { "mode": "effort", "efforts": [ @@ -38646,9 +39455,9 @@ "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.2, + "output": 1.15, + "cacheRead": 0.04, "cacheWrite": 0 }, "contextWindow": 256000, @@ -38662,8 +39471,7 @@ "high", "xhigh" ] - }, - "supportsComputerUse": false + } }, "stepfun/step-3.7-flash:free": { "id": "stepfun/step-3.7-flash:free", @@ -38671,9 +39479,10 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { "input": 0, @@ -38681,9 +39490,18 @@ "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 256000, - "maxTokens": 256000, - "supportsComputerUse": false + "contextWindow": 262144, + "maxTokens": 262144, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "switchpoint/router": { "id": "switchpoint/router", @@ -38732,23 +39550,32 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.14, + "output": 0.58, + "cacheRead": 0.035, "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 262144, - "supportsComputerUse": false + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "tencent/hy3-preview": { "id": "tencent/hy3-preview", - "name": "Hy3 Preview", + "name": "Hy3 preview", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -38757,9 +39584,9 @@ "text" ], "cost": { - "input": 0.066, - "output": 0.26, - "cacheRead": 0.029, + "input": 0.18, + "output": 0.6, + "cacheRead": 0.06, "cacheWrite": 0 }, "contextWindow": 262144, @@ -38802,7 +39629,7 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ "text" ], @@ -38813,9 +39640,17 @@ "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 262144, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "thedrummer/cydonia-24b-v4.1": { "id": "thedrummer/cydonia-24b-v4.1", @@ -38854,7 +39689,8 @@ "cacheWrite": 0 }, "contextWindow": 32768, - "maxTokens": 32768 + "maxTokens": 32768, + "supportsComputerUse": false }, "thedrummer/skyfall-36b-v2": { "id": "thedrummer/skyfall-36b-v2", @@ -38892,8 +39728,8 @@ "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 32768, - "maxTokens": 32768 + "contextWindow": 1024000, + "maxTokens": 1024000 }, "thinkingmachines/inkling": { "id": "thinkingmachines/inkling", @@ -38907,14 +39743,13 @@ "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.95, + "output": 4.05, + "cacheRead": 0.16, "cacheWrite": 0 }, - "contextWindow": 1000000, - "maxTokens": 65536, - "supportsComputerUse": false, + "contextWindow": 524288, + "maxTokens": 262144, "thinking": { "mode": "effort", "efforts": [ @@ -38932,19 +39767,29 @@ "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.45, + "output": 1.2, + "cacheRead": 0.1, "cacheWrite": 0 }, - "contextWindow": 1048576, - "maxTokens": 32768, - "supportsComputerUse": false + "contextWindow": 524288, + "maxTokens": 262144, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "tngtech/deepseek-r1t2-chimera": { "id": "tngtech/deepseek-r1t2-chimera", @@ -39000,11 +39845,11 @@ "cost": { "input": 0.15, "output": 0.6, - "cacheRead": 0, + "cacheRead": 0.015, "cacheWrite": 0 }, - "contextWindow": 128000, - "maxTokens": 32768, + "contextWindow": 131072, + "maxTokens": 131072, "thinking": { "mode": "effort", "efforts": [ @@ -39228,8 +40073,8 @@ "image" ], "cost": { - "input": 2, - "output": 6, + "input": 1.25, + "output": 2.5, "cacheRead": 0.2, "cacheWrite": 0 }, @@ -39361,9 +40206,9 @@ "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 2, + "output": 6, + "cacheRead": 0.3, "cacheWrite": 0 }, "contextWindow": 500000, @@ -39377,8 +40222,7 @@ "high", "xhigh" ] - }, - "supportsComputerUse": false + } }, "x-ai/grok-build-0.1": { "id": "x-ai/grok-build-0.1", @@ -39487,11 +40331,12 @@ "medium", "high" ] - } + }, + "supportsComputerUse": false }, "xiaomi/mimo-v2-omni": { "id": "xiaomi/mimo-v2-omni", - "name": "MiMo-V2-Omni", + "name": "MiMo V2 Omni", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -39515,7 +40360,8 @@ "medium", "high" ] - } + }, + "supportsComputerUse": false }, "xiaomi/mimo-v2-omni:free": { "id": "xiaomi/mimo-v2-omni:free", @@ -39540,7 +40386,7 @@ }, "xiaomi/mimo-v2-pro": { "id": "xiaomi/mimo-v2-pro", - "name": "MiMo-V2-Pro", + "name": "MiMo V2 Pro", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -39563,7 +40409,8 @@ "medium", "high" ] - } + }, + "supportsComputerUse": false }, "xiaomi/mimo-v2-pro:free": { "id": "xiaomi/mimo-v2-pro:free", @@ -39598,9 +40445,9 @@ "image" ], "cost": { - "input": 0.4, - "output": 2, - "cacheRead": 0.08, + "input": 0.14, + "output": 0.28, + "cacheRead": 0.0028, "cacheWrite": 0 }, "contextWindow": 1048576, @@ -39616,7 +40463,7 @@ }, "xiaomi/mimo-v2.5-pro": { "id": "xiaomi/mimo-v2.5-pro", - "name": "MiMo V2.5 Pro", + "name": "MiMo-V2.5-Pro", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -39625,9 +40472,9 @@ "text" ], "cost": { - "input": 1, - "output": 3, - "cacheRead": 0.2, + "input": 0.4, + "output": 1.5, + "cacheRead": 0.08, "cacheWrite": 0 }, "contextWindow": 1048576, @@ -39658,11 +40505,12 @@ "cacheWrite": 0 }, "contextWindow": 128000, - "maxTokens": 32768 + "maxTokens": 32768, + "supportsComputerUse": false }, "z-ai/glm-4.5": { "id": "z-ai/glm-4.5", - "name": "GLM 4.5", + "name": "GLM-4.5", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -39673,7 +40521,7 @@ "cost": { "input": 0.6, "output": 2.2, - "cacheRead": 0.175, + "cacheRead": 0.11, "cacheWrite": 0 }, "contextWindow": 131072, @@ -39691,7 +40539,7 @@ }, "z-ai/glm-4.5-air": { "id": "z-ai/glm-4.5-air", - "name": "GLM 4.5 Air", + "name": "GLM-4.5-Air", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -39720,7 +40568,7 @@ }, "z-ai/glm-4.5v": { "id": "z-ai/glm-4.5v", - "name": "GLM 4.5V", + "name": "GLM-4.5V", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -39750,7 +40598,7 @@ }, "z-ai/glm-4.6": { "id": "z-ai/glm-4.6", - "name": "GLM 4.6", + "name": "GLM-4.6", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -39759,13 +40607,13 @@ "text" ], "cost": { - "input": 0.39, - "output": 1.9, - "cacheRead": 0.175, + "input": 0.55, + "output": 2.2, + "cacheRead": 0.11, "cacheWrite": 0 }, - "contextWindow": 204800, - "maxTokens": 204800, + "contextWindow": 202752, + "maxTokens": 131072, "thinking": { "mode": "effort", "efforts": [ @@ -39800,7 +40648,7 @@ }, "z-ai/glm-4.6v": { "id": "z-ai/glm-4.6v", - "name": "GLM 4.6V", + "name": "GLM-4.6V", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -39812,11 +40660,11 @@ "cost": { "input": 0.3, "output": 0.9, - "cacheRead": 0, + "cacheRead": 0.055, "cacheWrite": 0 }, "contextWindow": 131072, - "maxTokens": 131072, + "maxTokens": 32768, "thinking": { "mode": "effort", "efforts": [ @@ -39830,7 +40678,7 @@ }, "z-ai/glm-4.7": { "id": "z-ai/glm-4.7", - "name": "GLM 4.7", + "name": "GLM-4.7", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -39839,13 +40687,13 @@ "text" ], "cost": { - "input": 0.38, - "output": 1.98, - "cacheRead": 0.2, + "input": 0.6, + "output": 2.2, + "cacheRead": 0.11, "cacheWrite": 0 }, "contextWindow": 202752, - "maxTokens": 65535, + "maxTokens": 131072, "thinking": { "mode": "effort", "efforts": [ @@ -39859,7 +40707,7 @@ }, "z-ai/glm-4.7-flash": { "id": "z-ai/glm-4.7-flash", - "name": "GLM 4.7 Flash", + "name": "GLM-4.7-Flash", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -39868,13 +40716,13 @@ "text" ], "cost": { - "input": 0.06, + "input": 0.07, "output": 0.4, "cacheRead": 0.01, "cacheWrite": 0 }, "contextWindow": 202752, - "maxTokens": 40551, + "maxTokens": 16384, "thinking": { "mode": "effort", "efforts": [ @@ -39888,7 +40736,7 @@ }, "z-ai/glm-5": { "id": "z-ai/glm-5", - "name": "GLM 5", + "name": "GLM-5", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -39897,12 +40745,12 @@ "text" ], "cost": { - "input": 0.72, - "output": 2.3, - "cacheRead": 0, + "input": 1, + "output": 3.2, + "cacheRead": 0.2, "cacheWrite": 0 }, - "contextWindow": 202752, + "contextWindow": 204800, "maxTokens": 131072, "thinking": { "mode": "effort", @@ -39917,7 +40765,7 @@ }, "z-ai/glm-5-turbo": { "id": "z-ai/glm-5-turbo", - "name": "GLM 5 Turbo", + "name": "GLM-5-Turbo", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -39946,7 +40794,7 @@ }, "z-ai/glm-5.1": { "id": "z-ai/glm-5.1", - "name": "GLM 5.1", + "name": "GLM-5.1", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -39955,9 +40803,9 @@ "text" ], "cost": { - "input": 1.26, - "output": 3.96, - "cacheRead": 0, + "input": 1.38, + "output": 4.4, + "cacheRead": 0.26, "cacheWrite": 0 }, "contextWindow": 202752, @@ -39984,9 +40832,9 @@ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 1.4, + "output": 4.4, + "cacheRead": 0.26, "cacheWrite": 0 }, "contextWindow": 1048576, @@ -40000,12 +40848,11 @@ "high", "max" ] - }, - "supportsComputerUse": false + } }, "z-ai/glm-5v-turbo": { "id": "z-ai/glm-5v-turbo", - "name": "GLM 5V Turbo", + "name": "GLM-5V-Turbo", "api": "openai-completions", "provider": "kilo", "baseUrl": "https://api.kilo.ai/api/gateway", @@ -42074,6 +42921,25 @@ }, "contextWindow": 128000, "maxTokens": 128000 + }, + "voxtral-small-latest": { + "id": "voxtral-small-latest", + "name": "Voxtral Small", + "api": "openai-completions", + "provider": "mistral", + "baseUrl": "https://api.mistral.ai/v1", + "reasoning": false, + "input": [ + "text" + ], + "cost": { + "input": 0.1, + "output": 0.3, + "cacheRead": 0, + "cacheWrite": 0 + }, + "contextWindow": 32000, + "maxTokens": 32000 } }, "moonshot": { @@ -42544,11 +43410,11 @@ }, "aion-labs/aion-2.0": { "id": "aion-labs/aion-2.0", - "name": "aion-labs/aion-2.0", + "name": "Aion-2.0", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": false, + "reasoning": true, "input": [ "text" ], @@ -42558,10 +43424,20 @@ "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": null, - "maxTokens": null, + "contextWindow": 131072, + "maxTokens": 32768, "supportsComputerUse": false, - "supportsComputerUseConfig": false + "supportsComputerUseConfig": false, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "aion-labs/aion-2.5": { "id": "aion-labs/aion-2.5", @@ -42586,11 +43462,11 @@ }, "aion-labs/aion-3.0": { "id": "aion-labs/aion-3.0", - "name": "aion-labs/aion-3.0", + "name": "Aion-3.0", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": false, + "reasoning": true, "input": [ "text" ], @@ -42600,18 +43476,28 @@ "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": null, - "maxTokens": null, + "contextWindow": 131072, + "maxTokens": 32768, "supportsComputerUse": false, - "supportsComputerUseConfig": false + "supportsComputerUseConfig": false, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "aion-labs/aion-3.0-mini": { "id": "aion-labs/aion-3.0-mini", - "name": "aion-labs/aion-3.0-mini", + "name": "Aion-3.0-Mini", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": false, + "reasoning": true, "input": [ "text" ], @@ -42621,10 +43507,20 @@ "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": null, - "maxTokens": null, + "contextWindow": 131072, + "maxTokens": 32768, "supportsComputerUse": false, - "supportsComputerUseConfig": false + "supportsComputerUseConfig": false, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "aion-labs/aion-rp-llama-3.1-8b": { "id": "aion-labs/aion-rp-llama-3.1-8b", @@ -42979,30 +43875,37 @@ "high", "xhigh" ] - }, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + } }, "anthropic/claude-fable-latest": { "id": "anthropic/claude-fable-latest", - "name": "anthropic/claude-fable-latest", + "name": "Claude Fable Latest", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 10, + "output": 50, + "cacheRead": 1, + "cacheWrite": 12.5 }, - "contextWindow": null, - "maxTokens": null, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "contextWindow": 1000000, + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "anthropic/claude-haiku-latest": { "id": "anthropic/claude-haiku-latest", @@ -43046,9 +43949,9 @@ "image" ], "cost": { - "input": 4.998, - "output": 25.007, - "cacheRead": 0, + "input": 5, + "output": 25, + "cacheRead": 0.5, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -43076,9 +43979,9 @@ "image" ], "cost": { - "input": 4.998, - "output": 25.007, - "cacheRead": 0, + "input": 5, + "output": 25, + "cacheRead": 0.5, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -43106,9 +44009,9 @@ "image" ], "cost": { - "input": 4.998, - "output": 25.007, - "cacheRead": 0, + "input": 5, + "output": 25, + "cacheRead": 0.5, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -43136,9 +44039,9 @@ "image" ], "cost": { - "input": 4.998, - "output": 25.007, - "cacheRead": 0, + "input": 5, + "output": 25, + "cacheRead": 0.5, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -43166,9 +44069,9 @@ "image" ], "cost": { - "input": 4.998, - "output": 25.007, - "cacheRead": 0, + "input": 5, + "output": 25, + "cacheRead": 0.5, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -43196,9 +44099,9 @@ "image" ], "cost": { - "input": 4.998, - "output": 25.007, - "cacheRead": 0.4998, + "input": 5, + "output": 25, + "cacheRead": 0.5, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -43226,9 +44129,9 @@ "image" ], "cost": { - "input": 4.998, - "output": 25.007, - "cacheRead": 0.4998, + "input": 5, + "output": 25, + "cacheRead": 0.5, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -43256,9 +44159,9 @@ "image" ], "cost": { - "input": 4.998, - "output": 25.007, - "cacheRead": 0.4998, + "input": 5, + "output": 25, + "cacheRead": 0.5, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -43286,9 +44189,9 @@ "image" ], "cost": { - "input": 4.998, - "output": 25.007, - "cacheRead": 0.4998, + "input": 5, + "output": 25, + "cacheRead": 0.5, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -43316,14 +44219,13 @@ "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 5, + "output": 25, + "cacheRead": 0.5, + "cacheWrite": 6.25 }, "contextWindow": 1000000, "maxTokens": 128000, - "supportsComputerUse": false, "thinking": { "mode": "effort", "efforts": [ @@ -43333,8 +44235,7 @@ "high", "xhigh" ] - }, - "supportsComputerUseConfig": false + } }, "anthropic/claude-opus-latest": { "id": "anthropic/claude-opus-latest", @@ -43348,10 +44249,10 @@ "image" ], "cost": { - "input": 4.998, - "output": 25.007, - "cacheRead": 0.4998, - "cacheWrite": 0 + "input": 5, + "output": 25, + "cacheRead": 0.5, + "cacheWrite": 6.25 }, "contextWindow": 1000000, "maxTokens": 128000, @@ -43372,19 +44273,29 @@ "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": false, + "reasoning": true, "input": [ "text", "image" ], "cost": { - "input": 2.992, - "output": 14.993999999999998, - "cacheRead": 0, + "input": 3, + "output": 15, + "cacheRead": 0.3, "cacheWrite": 0 }, "contextWindow": 1000000, - "maxTokens": 128000 + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "anthropic/claude-sonnet-4.6:thinking": { "id": "anthropic/claude-sonnet-4.6:thinking", @@ -43398,9 +44309,9 @@ "image" ], "cost": { - "input": 2.992, - "output": 14.993999999999998, - "cacheRead": 0, + "input": 3, + "output": 15, + "cacheRead": 0.3, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -43422,16 +44333,36 @@ "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": false, + "input": [ + "text", + "image" + ], + "cost": { + "input": 2, + "output": 10, + "cacheRead": 0.2, + "cacheWrite": 2.5 + }, + "contextWindow": 1000000, + "maxTokens": 128000 + }, + "anthropic/claude-sonnet-5:thinking": { + "id": "anthropic/claude-sonnet-5:thinking", + "name": "Claude Sonnet 5 Thinking", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", "reasoning": true, "input": [ "text", "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 2, + "output": 10, + "cacheRead": 0.2, + "cacheWrite": 2.5 }, "contextWindow": 1000000, "maxTokens": 128000, @@ -43444,9 +44375,7 @@ "high", "xhigh" ] - }, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + } }, "anthropic/claude-sonnet-latest": { "id": "anthropic/claude-sonnet-latest", @@ -43460,10 +44389,10 @@ "image" ], "cost": { - "input": 2.992, - "output": 14.994, - "cacheRead": 0.2992, - "cacheWrite": 0 + "input": 2, + "output": 10, + "cacheRead": 0.2, + "cacheWrite": 2.5 }, "contextWindow": 1000000, "maxTokens": 128000, @@ -43707,9 +44636,9 @@ "image" ], "cost": { - "input": 2.499, - "output": 9.996, - "cacheRead": 0, + "input": 2.5, + "output": 10, + "cacheRead": 1.25, "cacheWrite": 0 }, "contextWindow": 128000, @@ -43727,9 +44656,9 @@ "image" ], "cost": { - "input": 0.1496, - "output": 0.595, - "cacheRead": 0, + "input": 0.15, + "output": 0.6, + "cacheRead": 0.075, "cacheWrite": 0 }, "contextWindow": 128000, @@ -44342,26 +45271,6 @@ "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": false, - "input": [ - "text", - "image" - ], - "cost": { - "input": 1, - "output": 5, - "cacheRead": 0, - "cacheWrite": 0 - }, - "contextWindow": 200000, - "maxTokens": 64000 - }, - "claude-haiku-4-5-20251001-thinking": { - "id": "claude-haiku-4-5-20251001-thinking", - "name": "Claude Haiku 4.5 Thinking", - "api": "openai-completions", - "provider": "nanogpt", - "baseUrl": "https://nano-gpt.com/api/v1", "reasoning": true, "input": [ "text", @@ -44384,7 +45293,14 @@ "high", "xhigh" ], - "requiresEffort": true + "effortRouting": { + "off": "claude-haiku-4-5-20251001", + "minimal": "claude-haiku-4-5-20251001-thinking", + "low": "claude-haiku-4-5-20251001-thinking", + "medium": "claude-haiku-4-5-20251001-thinking", + "high": "claude-haiku-4-5-20251001-thinking", + "xhigh": "claude-haiku-4-5-20251001-thinking" + } } }, "claude-opus-4-1-20250805": { @@ -44399,9 +45315,9 @@ "image" ], "cost": { - "input": 14.994, - "output": 75.004, - "cacheRead": 0, + "input": 15, + "output": 75, + "cacheRead": 1.5, "cacheWrite": 0 }, "contextWindow": 200000, @@ -44419,9 +45335,9 @@ "image" ], "cost": { - "input": 14.994, - "output": 75.004, - "cacheRead": 0, + "input": 15, + "output": 75, + "cacheRead": 1.5, "cacheWrite": 0 }, "contextWindow": 200000, @@ -44450,9 +45366,9 @@ "image" ], "cost": { - "input": 14.994, - "output": 75.004, - "cacheRead": 0, + "input": 15, + "output": 75, + "cacheRead": 1.5, "cacheWrite": 0 }, "contextWindow": 200000, @@ -44481,9 +45397,9 @@ "image" ], "cost": { - "input": 14.994, - "output": 75.004, - "cacheRead": 0, + "input": 15, + "output": 75, + "cacheRead": 1.5, "cacheWrite": 0 }, "contextWindow": 200000, @@ -44512,9 +45428,9 @@ "image" ], "cost": { - "input": 14.994, - "output": 75.004, - "cacheRead": 0, + "input": 15, + "output": 75, + "cacheRead": 1.5, "cacheWrite": 0 }, "contextWindow": 200000, @@ -44543,9 +45459,9 @@ "image" ], "cost": { - "input": 14.994, - "output": 75.004, - "cacheRead": 0, + "input": 15, + "output": 75, + "cacheRead": 1.5, "cacheWrite": 0 }, "contextWindow": 200000, @@ -44574,9 +45490,9 @@ "image" ], "cost": { - "input": 14.994, - "output": 75.004, - "cacheRead": 0, + "input": 15, + "output": 75, + "cacheRead": 1.5, "cacheWrite": 0 }, "contextWindow": 200000, @@ -44594,9 +45510,9 @@ "image" ], "cost": { - "input": 4.998, - "output": 25.007, - "cacheRead": 0, + "input": 5, + "output": 25, + "cacheRead": 0.5, "cacheWrite": 0 }, "contextWindow": 200000, @@ -44624,9 +45540,9 @@ "image" ], "cost": { - "input": 4.998, - "output": 25.007, - "cacheRead": 0, + "input": 5, + "output": 25, + "cacheRead": 0.5, "cacheWrite": 0 }, "contextWindow": 200000, @@ -44654,9 +45570,9 @@ "image" ], "cost": { - "input": 14.994, - "output": 75.004, - "cacheRead": 0, + "input": 15, + "output": 75, + "cacheRead": 1.5, "cacheWrite": 0 }, "contextWindow": 200000, @@ -44685,9 +45601,9 @@ "image" ], "cost": { - "input": 14.994, - "output": 75.004, - "cacheRead": 0, + "input": 15, + "output": 75, + "cacheRead": 1.5, "cacheWrite": 0 }, "contextWindow": 200000, @@ -44716,9 +45632,9 @@ "image" ], "cost": { - "input": 14.994, - "output": 75.004, - "cacheRead": 0, + "input": 15, + "output": 75, + "cacheRead": 1.5, "cacheWrite": 0 }, "contextWindow": 200000, @@ -44747,9 +45663,9 @@ "image" ], "cost": { - "input": 14.994, - "output": 75.004, - "cacheRead": 0, + "input": 15, + "output": 75, + "cacheRead": 1.5, "cacheWrite": 0 }, "contextWindow": 200000, @@ -44778,9 +45694,9 @@ "image" ], "cost": { - "input": 14.994, - "output": 75.004, - "cacheRead": 0, + "input": 15, + "output": 75, + "cacheRead": 1.5, "cacheWrite": 0 }, "contextWindow": 200000, @@ -44809,9 +45725,9 @@ "image" ], "cost": { - "input": 2.992, - "output": 14.994, - "cacheRead": 0, + "input": 3, + "output": 15, + "cacheRead": 0.3, "cacheWrite": 0 }, "contextWindow": 200000, @@ -44829,9 +45745,9 @@ "image" ], "cost": { - "input": 2.992, - "output": 14.994, - "cacheRead": 0, + "input": 3, + "output": 15, + "cacheRead": 0.3, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -44867,9 +45783,9 @@ "image" ], "cost": { - "input": 2.992, - "output": 14.994, - "cacheRead": 0, + "input": 3, + "output": 15, + "cacheRead": 0.3, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -44898,9 +45814,9 @@ "image" ], "cost": { - "input": 2.992, - "output": 14.994, - "cacheRead": 0, + "input": 3, + "output": 15, + "cacheRead": 0.3, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -44929,9 +45845,9 @@ "image" ], "cost": { - "input": 2.992, - "output": 14.994, - "cacheRead": 0, + "input": 3, + "output": 15, + "cacheRead": 0.3, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -44960,9 +45876,9 @@ "image" ], "cost": { - "input": 2.992, - "output": 14.994, - "cacheRead": 0, + "input": 3, + "output": 15, + "cacheRead": 0.3, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -44991,9 +45907,9 @@ "image" ], "cost": { - "input": 2.992, - "output": 14.994, - "cacheRead": 0, + "input": 3, + "output": 15, + "cacheRead": 0.3, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -45022,9 +45938,9 @@ "image" ], "cost": { - "input": 4.998, - "output": 25.007, - "cacheRead": 0, + "input": 5, + "output": 25, + "cacheRead": 2.5, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -45055,7 +45971,7 @@ "input": 0.25, "output": 1.5, "cacheRead": 0.025, - "cacheWrite": 0 + "cacheWrite": 0.08333 }, "contextWindow": 1048576, "maxTokens": 65536, @@ -45081,9 +45997,9 @@ "text" ], "cost": { - "input": 0.3, - "output": 1.2, - "cacheRead": 0, + "input": 0.315, + "output": 1.26, + "cacheRead": 0.1575, "cacheWrite": 0 }, "contextWindow": 204800, @@ -45225,7 +46141,7 @@ }, "crofai/greg-2-super": { "id": "crofai/greg-2-super", - "name": "crofai/greg-2-super", + "name": "Greg 2 Super", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -45234,19 +46150,17 @@ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 1.5, + "output": 5, + "cacheRead": 0.25, "cacheWrite": 0 }, - "contextWindow": null, - "maxTokens": null, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "contextWindow": 229376, + "maxTokens": 229376 }, "crofai/greg-2-ultra": { "id": "crofai/greg-2-ultra", - "name": "crofai/greg-2-ultra", + "name": "Greg 2 Ultra", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -45255,15 +46169,13 @@ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 3, + "output": 10, + "cacheRead": 0.5, "cacheWrite": 0 }, - "contextWindow": null, - "maxTokens": null, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "contextWindow": 229376, + "maxTokens": 229376 }, "CrucibleLab/L3.3-70B-Loki-V2.0": { "id": "CrucibleLab/L3.3-70B-Loki-V2.0", @@ -45411,7 +46323,7 @@ "cost": { "input": 0.25, "output": 0.7, - "cacheRead": 0, + "cacheRead": 0.125, "cacheWrite": 0 }, "contextWindow": 128000, @@ -45430,7 +46342,7 @@ "cost": { "input": 0.25, "output": 0.7, - "cacheRead": 0, + "cacheRead": 0.125, "cacheWrite": 0 }, "contextWindow": 128000, @@ -45488,9 +46400,9 @@ "text" ], "cost": { - "input": 0.25, - "output": 0.7, - "cacheRead": 0, + "input": 0.1, + "output": 0.425, + "cacheRead": 0.05, "cacheWrite": 0 }, "contextWindow": 128000, @@ -45507,9 +46419,9 @@ "text" ], "cost": { - "input": 0.25, - "output": 0.7, - "cacheRead": 0, + "input": 0.1, + "output": 0.425, + "cacheRead": 0.05, "cacheWrite": 0 }, "contextWindow": 128000, @@ -45631,9 +46543,9 @@ "text" ], "cost": { - "input": 0.25, - "output": 0.7, - "cacheRead": 0, + "input": 0.2, + "output": 0.77, + "cacheRead": 0.135, "cacheWrite": 0 }, "contextWindow": 128000, @@ -45697,9 +46609,9 @@ "text" ], "cost": { - "input": 0.27999999999999997, - "output": 0.42000000000000004, - "cacheRead": 0, + "input": 0.28, + "output": 0.42, + "cacheRead": 0.14, "cacheWrite": 0 }, "contextWindow": 163000, @@ -45737,9 +46649,9 @@ "text" ], "cost": { - "input": 0.27999999999999997, - "output": 0.42000000000000004, - "cacheRead": 0, + "input": 0.28, + "output": 0.42, + "cacheRead": 0.14, "cacheWrite": 0 }, "contextWindow": 163000, @@ -45763,9 +46675,9 @@ "text" ], "cost": { - "input": 0.14, - "output": 0.28, - "cacheRead": 0.028, + "input": 0.07, + "output": 0.14, + "cacheRead": 0.014, "cacheWrite": 0 }, "contextWindow": 1048576, @@ -45773,6 +46685,7 @@ "thinking": { "mode": "effort", "efforts": [ + "low", "high", "max" ] @@ -45780,7 +46693,7 @@ }, "deepseek/deepseek-v4-flash-0731": { "id": "deepseek/deepseek-v4-flash-0731", - "name": "deepseek/deepseek-v4-flash-0731", + "name": "DeepSeek V4 Flash 0731", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -45789,18 +46702,125 @@ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.14, + "output": 0.28, + "cacheRead": 0.014, + "cacheWrite": 0 + }, + "contextWindow": 1000000, + "maxTokens": 131072, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "high", + "max" + ] + } + }, + "deepseek/deepseek-v4-flash-0731-cheaper": { + "id": "deepseek/deepseek-v4-flash-0731-cheaper", + "name": "DeepSeek V4 Flash 0731 Cheaper", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.14, + "output": 0.28, + "cacheRead": 0.014, "cacheWrite": 0 }, "contextWindow": 1048576, "maxTokens": 384000, - "supportsComputerUse": false, - "supportsComputerUseConfig": false, "thinking": { "mode": "effort", "efforts": [ + "low", + "high", + "max" + ] + } + }, + "deepseek/deepseek-v4-flash-0731-cheaper:thinking": { + "id": "deepseek/deepseek-v4-flash-0731-cheaper:thinking", + "name": "DeepSeek V4 Flash 0731 Cheaper (Thinking)", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.14, + "output": 0.28, + "cacheRead": 0.014, + "cacheWrite": 0 + }, + "contextWindow": 1048576, + "maxTokens": 384000, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "high", + "max" + ] + } + }, + "deepseek/deepseek-v4-flash-0731:thinking": { + "id": "deepseek/deepseek-v4-flash-0731:thinking", + "name": "DeepSeek V4 Flash 0731 (Thinking)", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.14, + "output": 0.28, + "cacheRead": 0.014, + "cacheWrite": 0 + }, + "contextWindow": 1000000, + "maxTokens": 131072, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "high", + "max" + ] + } + }, + "deepseek/deepseek-v4-flash-latest": { + "id": "deepseek/deepseek-v4-flash-latest", + "name": "DeepSeek V4 Flash Latest", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.14, + "output": 0.28, + "cacheRead": 0.014, + "cacheWrite": 0 + }, + "contextWindow": 1048576, + "maxTokens": 384000, + "thinking": { + "mode": "effort", + "efforts": [ + "low", "high", "max" ] @@ -45817,9 +46837,9 @@ "text" ], "cost": { - "input": 0.14, - "output": 0.28, - "cacheRead": 0.028, + "input": 0.07, + "output": 0.14, + "cacheRead": 0.014, "cacheWrite": 0 }, "contextWindow": 1048576, @@ -45827,6 +46847,7 @@ "thinking": { "mode": "effort", "efforts": [ + "low", "high", "max" ] @@ -46995,7 +48016,7 @@ "cost": { "input": 0.1, "output": 0.4, - "cacheRead": 0, + "cacheRead": 0.01, "cacheWrite": 0 }, "contextWindow": 1048756, @@ -47114,7 +48135,7 @@ "cost": { "input": 0.3, "output": 2.5, - "cacheRead": 0, + "cacheRead": 0.03, "cacheWrite": 0 }, "contextWindow": 1048756, @@ -47487,6 +48508,26 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "gemma-4-12b-it": { + "id": "gemma-4-12b-it", + "name": "Gemma 4 12B Instruct", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": false, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.06, + "output": 0.3, + "cacheRead": 0.03, + "cacheWrite": 0 + }, + "contextWindow": 262144, + "maxTokens": 32768 + }, "Gemma-4-31B-Agares-v1": { "id": "Gemma-4-31B-Agares-v1", "name": "Gemma-4-31B-Agares-v1", @@ -48264,6 +49305,46 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "gemma-4-e2b-it": { + "id": "gemma-4-e2b-it", + "name": "Gemma 4 E2B Instruct", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": false, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.02, + "output": 0.1, + "cacheRead": 0.01, + "cacheWrite": 0 + }, + "contextWindow": 131072, + "maxTokens": 16384 + }, + "gemma-4-e4b-it": { + "id": "gemma-4-e4b-it", + "name": "Gemma 4 E4B Instruct", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": false, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.04, + "output": 0.2, + "cacheRead": 0.02, + "cacheWrite": 0 + }, + "contextWindow": 131072, + "maxTokens": 16384 + }, "glm-4": { "id": "glm-4", "name": "glm-4", @@ -48676,7 +49757,7 @@ "cost": { "input": 0.07, "output": 0.07, - "cacheRead": 0, + "cacheRead": 0.035, "cacheWrite": 0 }, "contextWindow": 32000, @@ -48695,7 +49776,7 @@ "cost": { "input": 0.7, "output": 0.7, - "cacheRead": 0, + "cacheRead": 0.35, "cacheWrite": 0 }, "contextWindow": 32000, @@ -48736,7 +49817,7 @@ "cost": { "input": 0.5, "output": 3, - "cacheRead": 0, + "cacheRead": 0.05, "cacheWrite": 0 }, "contextWindow": 1048756, @@ -48767,7 +49848,7 @@ "input": 0.25, "output": 1.5, "cacheRead": 0.025, - "cacheWrite": 0 + "cacheWrite": 0.08333 }, "contextWindow": 1048576, "maxTokens": 65536, @@ -48819,7 +49900,7 @@ "input": 2, "output": 12, "cacheRead": 0.2, - "cacheWrite": 0 + "cacheWrite": 0.375 }, "contextWindow": 1048756, "maxTokens": 65536, @@ -48847,7 +49928,7 @@ "input": 2, "output": 12, "cacheRead": 0.2, - "cacheWrite": 0 + "cacheWrite": 0.375 }, "contextWindow": 1048756, "maxTokens": 65536, @@ -48875,7 +49956,7 @@ "input": 2, "output": 12, "cacheRead": 0.2, - "cacheWrite": 0 + "cacheWrite": 0.375 }, "contextWindow": 1048756, "maxTokens": 65536, @@ -48903,7 +49984,7 @@ "input": 2, "output": 12, "cacheRead": 0.2, - "cacheWrite": 0 + "cacheWrite": 0.375 }, "contextWindow": 1048756, "maxTokens": 65536, @@ -48948,45 +50029,63 @@ }, "google/gemini-3.5-flash-lite": { "id": "google/gemini-3.5-flash-lite", - "name": "google/gemini-3.5-flash-lite", + "name": "Gemini 3.5 Flash Lite", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 0.3, + "output": 2.5, + "cacheRead": 0.03, + "cacheWrite": 0.08333 }, "contextWindow": 1048576, "maxTokens": 65536, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ], + "requiresEffort": true + } }, "google/gemini-3.6-flash": { "id": "google/gemini-3.6-flash", - "name": "google/gemini-3.6-flash", + "name": "Gemini 3.6 Flash", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 1.5, + "output": 7.5, + "cacheRead": 0.15, "cacheWrite": 0 }, "contextWindow": 1048576, "maxTokens": 65536, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ], + "requiresEffort": true + } }, "google/gemini-flash-1.5": { "id": "google/gemini-flash-1.5", @@ -49051,10 +50150,10 @@ "image" ], "cost": { - "input": 0.25, - "output": 1.5, - "cacheRead": 0.025, - "cacheWrite": 0 + "input": 0.3, + "output": 2.5, + "cacheRead": 0.03, + "cacheWrite": 0.08333 }, "contextWindow": 1048576, "maxTokens": 65536, @@ -49084,7 +50183,7 @@ "input": 2, "output": 12, "cacheRead": 0.2, - "cacheWrite": 0 + "cacheWrite": 0.375 }, "contextWindow": 1048756, "maxTokens": 65536, @@ -49101,7 +50200,7 @@ }, "google/gemma-4-26b-a4b-it": { "id": "google/gemma-4-26b-a4b-it", - "name": "Gemma 4 26B A4B", + "name": "Gemma 4 26B A4B IT", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -49133,7 +50232,7 @@ }, "google/gemma-4-31b-it": { "id": "google/gemma-4-31b-it", - "name": "Gemma 4 31B", + "name": "Gemma 4 31B IT", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -49280,9 +50379,9 @@ "image" ], "cost": { - "input": 4.998, - "output": 25.007, - "cacheRead": 0, + "input": 5, + "output": 25, + "cacheRead": 2.5, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -49313,7 +50412,7 @@ "input": 0.25, "output": 1.5, "cacheRead": 0.025, - "cacheWrite": 0 + "cacheWrite": 0.08333 }, "contextWindow": 1048576, "maxTokens": 65536, @@ -49339,9 +50438,9 @@ "text" ], "cost": { - "input": 0.3, - "output": 1.2, - "cacheRead": 0, + "input": 0.315, + "output": 1.26, + "cacheRead": 0.1575, "cacheWrite": 0 }, "contextWindow": 204800, @@ -49371,7 +50470,7 @@ "cost": { "input": 0.25, "output": 1.8, - "cacheRead": 0, + "cacheRead": 0.125, "cacheWrite": 0 }, "contextWindow": 65536, @@ -49401,7 +50500,7 @@ "cost": { "input": 0.25, "output": 1.8, - "cacheRead": 0, + "cacheRead": 0.125, "cacheWrite": 0 }, "contextWindow": 65536, @@ -49613,9 +50712,9 @@ "text" ], "cost": { - "input": 0.08, - "output": 0.24, - "cacheRead": 0, + "input": 0.1, + "output": 0.3, + "cacheRead": 0.02, "cacheWrite": 0 }, "contextWindow": 262144, @@ -49623,7 +50722,26 @@ }, "inclusionai/ling-3.0-flash": { "id": "inclusionai/ling-3.0-flash", - "name": "inclusionai/ling-3.0-flash", + "name": "Ling 3.0 Flash", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": false, + "input": [ + "text" + ], + "cost": { + "input": 0.075, + "output": 0.22, + "cacheRead": 0.015, + "cacheWrite": 0 + }, + "contextWindow": 262144, + "maxTokens": 32768 + }, + "inclusionai/ling-3.0-flash:thinking": { + "id": "inclusionai/ling-3.0-flash:thinking", + "name": "Ling 3.0 Flash Thinking", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -49632,9 +50750,9 @@ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.075, + "output": 0.22, + "cacheRead": 0.015, "cacheWrite": 0 }, "contextWindow": 262144, @@ -49648,9 +50766,7 @@ "high", "xhigh" ] - }, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + } }, "inclusionai/ring-2.6-1t": { "id": "inclusionai/ring-2.6-1t", @@ -49663,9 +50779,9 @@ "text" ], "cost": { - "input": 1, - "output": 3, - "cacheRead": 0, + "input": 0.3, + "output": 2.5, + "cacheRead": 0.06, "cacheWrite": 0 }, "contextWindow": 262144, @@ -49998,7 +51114,7 @@ }, "kwaipilot/kat-coder-air-v2.5": { "id": "kwaipilot/kat-coder-air-v2.5", - "name": "kwaipilot/kat-coder-air-v2.5", + "name": "KAT Coder Air V2.5", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -50007,15 +51123,13 @@ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.15, + "output": 0.6, + "cacheRead": 0.03, "cacheWrite": 0 }, - "contextWindow": null, - "maxTokens": null, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "contextWindow": 256000, + "maxTokens": 80000 }, "kwaipilot/kat-coder-pro-v2": { "id": "kwaipilot/kat-coder-pro-v2", @@ -50040,7 +51154,7 @@ }, "kwaipilot/kat-coder-pro-v2.5": { "id": "kwaipilot/kat-coder-pro-v2.5", - "name": "kwaipilot/kat-coder-pro-v2.5", + "name": "KAT Coder Pro V2.5", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -50049,15 +51163,13 @@ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.74, + "output": 2.96, + "cacheRead": 0.15, "cacheWrite": 0 }, - "contextWindow": null, - "maxTokens": null, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "contextWindow": 256000, + "maxTokens": 80000 }, "LatitudeGames/Wayfarer-Large-70B-Llama-3.3": { "id": "LatitudeGames/Wayfarer-Large-70B-Llama-3.3", @@ -51132,7 +52244,26 @@ }, "longcat-2.0": { "id": "longcat-2.0", - "name": "longcat-2.0", + "name": "LongCat 2.0", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": false, + "input": [ + "text" + ], + "cost": { + "input": 0.75, + "output": 3, + "cacheRead": 0.015, + "cacheWrite": 0 + }, + "contextWindow": 1048756, + "maxTokens": 262144 + }, + "longcat-2.0:thinking": { + "id": "longcat-2.0:thinking", + "name": "LongCat 2.0 Thinking", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -51141,13 +52272,13 @@ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.75, + "output": 3, + "cacheRead": 0.015, "cacheWrite": 0 }, - "contextWindow": null, - "maxTokens": null, + "contextWindow": 1048756, + "maxTokens": 262144, "thinking": { "mode": "effort", "efforts": [ @@ -51157,9 +52288,7 @@ "high", "xhigh" ] - }, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + } }, "Magistral-Small-2506": { "id": "Magistral-Small-2506", @@ -51434,7 +52563,7 @@ "cost": { "input": 0.05, "output": 0.23, - "cacheRead": 0, + "cacheRead": 0.025, "cacheWrite": 0 }, "contextWindow": 131072, @@ -51474,7 +52603,7 @@ "cost": { "input": 0.085, "output": 0.46, - "cacheRead": 0, + "cacheRead": 0.0425, "cacheWrite": 0 }, "contextWindow": 328000, @@ -51482,13 +52611,14 @@ }, "meta/muse-spark-1.1": { "id": "meta/muse-spark-1.1", - "name": "meta/muse-spark-1.1", + "name": "Muse Spark 1.1", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { "input": 0, @@ -51499,7 +52629,77 @@ "contextWindow": 1048576, "maxTokens": 1048576, "supportsComputerUse": false, - "supportsComputerUseConfig": false + "supportsComputerUseConfig": false, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } + }, + "meta/muse-spark-1.2": { + "id": "meta/muse-spark-1.2", + "name": "Muse Spark 1.2", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 1.25, + "output": 4.25, + "cacheRead": 0.15, + "cacheWrite": 0 + }, + "contextWindow": 1000000, + "maxTokens": 65536, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } + }, + "meta/muse-spark-1.2-contributor": { + "id": "meta/muse-spark-1.2-contributor", + "name": "Muse Spark 1.2 Contributor (Data Used for Training)", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.1, + "output": 0.2, + "cacheRead": 0.002, + "cacheWrite": 0 + }, + "contextWindow": 1000000, + "maxTokens": 65536, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "microsoft/MAI-DS-R1-FP8": { "id": "microsoft/MAI-DS-R1-FP8", @@ -51682,7 +52882,7 @@ "cost": { "input": 0.33, "output": 1.32, - "cacheRead": 0, + "cacheRead": 0.165, "cacheWrite": 0 }, "contextWindow": 200000, @@ -51710,7 +52910,7 @@ "cost": { "input": 0.3, "output": 1.2, - "cacheRead": 0, + "cacheRead": 0.15, "cacheWrite": 0 }, "contextWindow": 204800, @@ -51736,9 +52936,9 @@ "text" ], "cost": { - "input": 0.3, - "output": 1.2, - "cacheRead": 0, + "input": 0.315, + "output": 1.26, + "cacheRead": 0.1575, "cacheWrite": 0 }, "contextWindow": 204800, @@ -51766,7 +52966,7 @@ "cost": { "input": 0.6, "output": 2.4, - "cacheRead": 0, + "cacheRead": 0.3, "cacheWrite": 0 }, "contextWindow": 204800, @@ -51928,7 +53128,7 @@ "cost": { "input": 0.4, "output": 2, - "cacheRead": 0, + "cacheRead": 0.2, "cacheWrite": 0 }, "contextWindow": 262144, @@ -51947,7 +53147,7 @@ "cost": { "input": 0.3, "output": 0.9, - "cacheRead": 0, + "cacheRead": 0.15, "cacheWrite": 0 }, "contextWindow": 256000, @@ -52009,7 +53209,7 @@ "cost": { "input": 1.5, "output": 7.5, - "cacheRead": 0, + "cacheRead": 0.75, "cacheWrite": 0 }, "contextWindow": 256000, @@ -52039,7 +53239,7 @@ "cost": { "input": 1.5, "output": 7.5, - "cacheRead": 0, + "cacheRead": 0.75, "cacheWrite": 0 }, "contextWindow": 256000, @@ -52401,7 +53601,7 @@ "cost": { "input": 0.4, "output": 1.4, - "cacheRead": 0, + "cacheRead": 0.2, "cacheWrite": 0 }, "contextWindow": 262144, @@ -52431,7 +53631,7 @@ "cost": { "input": 0.4, "output": 1.4, - "cacheRead": 0, + "cacheRead": 0.2, "cacheWrite": 0 }, "contextWindow": 262144, @@ -52584,9 +53784,9 @@ "text" ], "cost": { - "input": 0.1, - "output": 2, - "cacheRead": 0, + "input": 0.4, + "output": 1.8, + "cacheRead": 0.2, "cacheWrite": 0 }, "contextWindow": 256000, @@ -52603,9 +53803,9 @@ "text" ], "cost": { - "input": 0.1, - "output": 2, - "cacheRead": 0, + "input": 0.4, + "output": 1.8, + "cacheRead": 0.2, "cacheWrite": 0 }, "contextWindow": 128000, @@ -52623,8 +53823,8 @@ ], "cost": { "input": 0.4, - "output": 2, - "cacheRead": 0, + "output": 1.8, + "cacheRead": 0.2, "cacheWrite": 0 }, "contextWindow": 256000, @@ -52636,18 +53836,29 @@ "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": false, + "reasoning": true, "input": [ "text" ], "cost": { - "input": 0.3, - "output": 1.2, - "cacheRead": 0, + "input": 0.6, + "output": 2.5, + "cacheRead": 0.15, "cacheWrite": 0 }, - "contextWindow": 256000, - "maxTokens": 262144 + "contextWindow": 262144, + "maxTokens": 98304, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ], + "requiresEffort": true + } }, "moonshotai/kimi-k2-thinking-original": { "id": "moonshotai/kimi-k2-thinking-original", @@ -52705,7 +53916,7 @@ "cost": { "input": 0.3, "output": 1.9, - "cacheRead": 0, + "cacheRead": 0.15, "cacheWrite": 0 }, "contextWindow": 256000, @@ -52725,7 +53936,7 @@ "cost": { "input": 0.3, "output": 1.9, - "cacheRead": 0, + "cacheRead": 0.15, "cacheWrite": 0 }, "contextWindow": 256000, @@ -52753,9 +53964,9 @@ "image" ], "cost": { - "input": 0.53, - "output": 2.73, - "cacheRead": 0, + "input": 0.5, + "output": 2.6, + "cacheRead": 0.125, "cacheWrite": 0 }, "contextWindow": 256000, @@ -52773,9 +53984,9 @@ "image" ], "cost": { - "input": 0.53, - "output": 2.73, - "cacheRead": 0, + "input": 0.5, + "output": 2.6, + "cacheRead": 0.125, "cacheWrite": 0 }, "contextWindow": 256000, @@ -52803,13 +54014,13 @@ "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.95, + "output": 4, + "cacheRead": 0.19, "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 262144, + "maxTokens": 65536, "thinking": { "mode": "effort", "efforts": [ @@ -52819,30 +54030,37 @@ "high", "xhigh" ] - }, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + } }, "moonshotai/kimi-k2.7-code-highspeed": { "id": "moonshotai/kimi-k2.7-code-highspeed", - "name": "moonshotai/kimi-k2.7-code-highspeed", + "name": "Kimi K2.7 Code High-Speed", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 1.9, + "output": 8, + "cacheRead": 0.32, "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 262144, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "maxTokens": 65536, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "moonshotai/kimi-k3": { "id": "moonshotai/kimi-k3", @@ -52856,13 +54074,13 @@ "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 2.5, + "output": 13.5, + "cacheRead": 0.25, "cacheWrite": 0 }, "contextWindow": 1048576, - "maxTokens": 131072, + "maxTokens": 1048576, "thinking": { "mode": "effort", "efforts": [ @@ -52875,9 +54093,7 @@ "max": "max" }, "requiresEffort": true - }, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + } }, "moonshotai/kimi-latest": { "id": "moonshotai/kimi-latest", @@ -52891,13 +54107,13 @@ "image" ], "cost": { - "input": 0.5, - "output": 2.6, - "cacheRead": 0.125, + "input": 2.5, + "output": 13.5, + "cacheRead": 0.25, "cacheWrite": 0 }, - "contextWindow": 256000, - "maxTokens": 65536, + "contextWindow": 1048576, + "maxTokens": 1048576, "thinking": { "mode": "effort", "efforts": [ @@ -53140,45 +54356,63 @@ }, "nex-agi/nex-n2-mini": { "id": "nex-agi/nex-n2-mini", - "name": "nex-agi/nex-n2-mini", + "name": "Nex N2 Mini", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.025, + "output": 0.1, + "cacheRead": 0.0025, "cacheWrite": 0 }, "contextWindow": 262144, "maxTokens": 262144, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "nex-agi/nex-n2-pro": { "id": "nex-agi/nex-n2-pro", - "name": "nex-agi/nex-n2-pro", + "name": "Nex N2 Pro", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.5, + "output": 2.5, + "cacheRead": 0.25, "cacheWrite": 0 }, "contextWindow": 262144, "maxTokens": 262144, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "nothingiisreal/L3.1-70B-Celeste-V0.1-BF16": { "id": "nothingiisreal/L3.1-70B-Celeste-V0.1-BF16", @@ -53434,7 +54668,7 @@ "cost": { "input": 0.105, "output": 0.42, - "cacheRead": 0, + "cacheRead": 0.0525, "cacheWrite": 0 }, "contextWindow": 256000, @@ -53464,7 +54698,7 @@ "cost": { "input": 0.05, "output": 0.25, - "cacheRead": 0, + "cacheRead": 0.025, "cacheWrite": 0 }, "contextWindow": 262144, @@ -53493,7 +54727,7 @@ "cost": { "input": 0.05, "output": 0.25, - "cacheRead": 0, + "cacheRead": 0.025, "cacheWrite": 0 }, "contextWindow": 262144, @@ -53511,7 +54745,7 @@ }, "nvidia/nemotron-3-ultra-550b-a55b": { "id": "nvidia/nemotron-3-ultra-550b-a55b", - "name": "Nemotron 3 Ultra 550B A55B", + "name": "Nvidia Nemotron 3 Ultra 550B", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -53520,9 +54754,9 @@ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.5, + "output": 2.5, + "cacheRead": 0.25, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -53536,9 +54770,36 @@ "high", "xhigh" ] + } + }, + "nvidia/nemotron-3-ultra-550b-a55b:thinking": { + "id": "nvidia/nemotron-3-ultra-550b-a55b:thinking", + "name": "Nvidia Nemotron 3 Ultra 550B Thinking", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.5, + "output": 2.5, + "cacheRead": 0.25, + "cacheWrite": 0 }, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "contextWindow": 1000000, + "maxTokens": 65536, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-TEE": { "id": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-TEE", @@ -53615,7 +54876,7 @@ }, "openai/gpt-3.5-turbo": { "id": "openai/gpt-3.5-turbo", - "name": "GPT-3.5 Turbo", + "name": "GPT-3.5-turbo", "api": "openai-completions", "baseUrl": "https://nano-gpt.com/api/v1", "provider": "nanogpt", @@ -53691,7 +54952,7 @@ "cost": { "input": 2, "output": 8, - "cacheRead": 0, + "cacheRead": 0.5, "cacheWrite": 0 }, "contextWindow": 1047576, @@ -53699,7 +54960,7 @@ }, "openai/gpt-4.1-mini": { "id": "openai/gpt-4.1-mini", - "name": "GPT-4.1 Mini", + "name": "GPT-4.1 mini", "api": "openai-completions", "baseUrl": "https://nano-gpt.com/api/v1", "provider": "nanogpt", @@ -53721,7 +54982,7 @@ }, "openai/gpt-4.1-nano": { "id": "openai/gpt-4.1-nano", - "name": "GPT-4.1 Nano", + "name": "GPT-4.1 nano", "api": "openai-completions", "baseUrl": "https://nano-gpt.com/api/v1", "provider": "nanogpt", @@ -53885,7 +55146,7 @@ "cost": { "input": 1.25, "output": 10, - "cacheRead": 0, + "cacheRead": 0.125, "cacheWrite": 0 }, "contextWindow": 400000, @@ -53925,12 +55186,11 @@ "id": "openai/gpt-5-codex", "name": "GPT-5 Codex", "api": "openai-completions", - "baseUrl": "https://nano-gpt.com/api/v1", "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", "reasoning": true, "input": [ - "text", - "image" + "text" ], "cost": { "input": 1.25, @@ -53939,7 +55199,7 @@ "cacheWrite": 0 }, "contextWindow": 272000, - "maxTokens": 64000, + "maxTokens": 32768, "thinking": { "mode": "effort", "efforts": [ @@ -53948,9 +55208,7 @@ "medium", "high" ] - }, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + } }, "openai/gpt-5-mini": { "id": "openai/gpt-5-mini", @@ -54059,7 +55317,7 @@ "cost": { "input": 1.25, "output": 10, - "cacheRead": 0, + "cacheRead": 0.125, "cacheWrite": 0 }, "contextWindow": 400000, @@ -54076,24 +55334,31 @@ }, "openai/gpt-5.1-2025-11-13": { "id": "openai/gpt-5.1-2025-11-13", - "name": "openai/gpt-5.1-2025-11-13", + "name": "GPT-5.1 (2025-11-13)", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": false, + "reasoning": true, "input": [ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 1.25, + "output": 10, + "cacheRead": 0.125, "cacheWrite": 0 }, - "contextWindow": 128000, - "maxTokens": 16384, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "contextWindow": 1000000, + "maxTokens": 32768, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + } }, "openai/gpt-5.1-chat": { "id": "openai/gpt-5.1-chat", @@ -54183,7 +55448,7 @@ "cost": { "input": 2.5, "output": 20, - "cacheRead": 0, + "cacheRead": 0.25, "cacheWrite": 0 }, "contextWindow": 272000, @@ -54200,7 +55465,7 @@ }, "openai/gpt-5.1-codex-mini": { "id": "openai/gpt-5.1-codex-mini", - "name": "GPT-5.1-Codex-Mini", + "name": "GPT-5.1 Codex mini", "api": "openai-completions", "baseUrl": "https://nano-gpt.com/api/v1", "provider": "nanogpt", @@ -54241,7 +55506,7 @@ "cost": { "input": 1.75, "output": 14, - "cacheRead": 0, + "cacheRead": 0.175, "cacheWrite": 0 }, "contextWindow": 400000, @@ -54292,7 +55557,7 @@ "cost": { "input": 1.75, "output": 14, - "cacheRead": 0, + "cacheRead": 0.175, "cacheWrite": 0 }, "contextWindow": 272000, @@ -54309,7 +55574,7 @@ }, "openai/gpt-5.2-pro": { "id": "openai/gpt-5.2-pro", - "name": "GPT 5.2 Pro", + "name": "GPT-5.2 Pro", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -54476,7 +55741,7 @@ }, "openai/gpt-5.4-pro": { "id": "openai/gpt-5.4-pro", - "name": "GPT 5.4 Pro", + "name": "GPT-5.4 Pro", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -54535,7 +55800,127 @@ }, "openai/gpt-5.6-luna": { "id": "openai/gpt-5.6-luna", - "name": "GPT-5.6 Luna", + "name": "GPT 5.6 Luna", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.1, + "output": 0.6, + "cacheRead": 0.01, + "cacheWrite": 0.125 + }, + "contextWindow": 1050000, + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "high", + "xhigh", + "max" + ] + } + }, + "openai/gpt-5.6-luna-pro": { + "id": "openai/gpt-5.6-luna-pro", + "name": "GPT 5.6 Luna Pro", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.1, + "output": 0.6, + "cacheRead": 0.01, + "cacheWrite": 0.125 + }, + "contextWindow": 1050000, + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "high", + "xhigh", + "max" + ] + } + }, + "openai/gpt-5.6-sol": { + "id": "openai/gpt-5.6-sol", + "name": "GPT 5.6 Sol", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 5, + "output": 30, + "cacheRead": 0.5, + "cacheWrite": 6.25 + }, + "contextWindow": 1050000, + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "high", + "xhigh", + "max" + ] + } + }, + "openai/gpt-5.6-sol-pro": { + "id": "openai/gpt-5.6-sol-pro", + "name": "GPT 5.6 Sol Pro", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 5, + "output": 30, + "cacheRead": 0.5, + "cacheWrite": 6.25 + }, + "contextWindow": 1050000, + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "high", + "xhigh", + "max" + ] + } + }, + "openai/gpt-5.6-terra": { + "id": "openai/gpt-5.6-terra", + "name": "GPT 5.6 Terra", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -54547,8 +55932,8 @@ "cost": { "input": 1, "output": 6, - "cacheRead": 0.09999999999999999, - "cacheWrite": 0 + "cacheRead": 0.1, + "cacheWrite": 1.25 }, "contextWindow": 1050000, "maxTokens": 128000, @@ -54561,136 +55946,37 @@ "xhigh", "max" ] - }, - "supportsComputerUse": false, - "supportsComputerUseConfig": false - }, - "openai/gpt-5.6-luna-pro": { - "id": "openai/gpt-5.6-luna-pro", - "name": "GPT-5.6 Luna Pro", - "api": "openai-completions", - "provider": "nanogpt", - "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": false, - "input": [ - "text" - ], - "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 - }, - "contextWindow": 1050000, - "maxTokens": 128000, - "supportsComputerUse": false, - "supportsComputerUseConfig": false - }, - "openai/gpt-5.6-sol": { - "id": "openai/gpt-5.6-sol", - "name": "GPT-5.6 Sol", - "api": "openai-completions", - "provider": "nanogpt", - "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": true, - "input": [ - "text", - "image" - ], - "cost": { - "input": 5, - "output": 30, - "cacheRead": 0.5, - "cacheWrite": 0 - }, - "contextWindow": 1050000, - "maxTokens": 128000, - "thinking": { - "mode": "effort", - "efforts": [ - "low", - "medium", - "high", - "xhigh", - "max" - ] - }, - "supportsComputerUse": false, - "supportsComputerUseConfig": false - }, - "openai/gpt-5.6-sol-pro": { - "id": "openai/gpt-5.6-sol-pro", - "name": "GPT-5.6 Sol Pro", - "api": "openai-completions", - "provider": "nanogpt", - "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": false, - "input": [ - "text" - ], - "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 - }, - "contextWindow": 1050000, - "maxTokens": 128000, - "supportsComputerUse": false, - "supportsComputerUseConfig": false - }, - "openai/gpt-5.6-terra": { - "id": "openai/gpt-5.6-terra", - "name": "GPT-5.6 Terra", - "api": "openai-completions", - "provider": "nanogpt", - "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": true, - "input": [ - "text", - "image" - ], - "cost": { - "input": 2.5, - "output": 15, - "cacheRead": 0.25, - "cacheWrite": 0 - }, - "contextWindow": 1050000, - "maxTokens": 128000, - "thinking": { - "mode": "effort", - "efforts": [ - "low", - "medium", - "high", - "xhigh", - "max" - ] - }, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + } }, "openai/gpt-5.6-terra-pro": { "id": "openai/gpt-5.6-terra-pro", - "name": "GPT-5.6 Terra Pro", + "name": "GPT 5.6 Terra Pro", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 1, + "output": 6, + "cacheRead": 0.1, + "cacheWrite": 1.25 }, "contextWindow": 1050000, "maxTokens": 128000, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "high", + "xhigh", + "max" + ] + } }, "openai/gpt-chat-latest": { "id": "openai/gpt-chat-latest", @@ -54698,7 +55984,7 @@ "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": false, + "reasoning": true, "input": [ "text", "image" @@ -54707,10 +55993,20 @@ "input": 5, "output": 30, "cacheRead": 0.5, - "cacheWrite": 0 + "cacheWrite": 6.25 }, - "contextWindow": 400000, - "maxTokens": 128000 + "contextWindow": 1050000, + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "openai/gpt-latest": { "id": "openai/gpt-latest", @@ -54727,9 +56023,9 @@ "input": 5, "output": 30, "cacheRead": 0.5, - "cacheWrite": 0 + "cacheWrite": 6.25 }, - "contextWindow": 1000000, + "contextWindow": 1050000, "maxTokens": 128000, "thinking": { "mode": "effort", @@ -54753,8 +56049,8 @@ "text" ], "cost": { - "input": 0.05, - "output": 0.25, + "input": 0.35, + "output": 0.75, "cacheRead": 0, "cacheWrite": 0 }, @@ -54981,7 +56277,7 @@ "cost": { "input": 1.1, "output": 4.4, - "cacheRead": 0, + "cacheRead": 0.55, "cacheWrite": 0 }, "contextWindow": 200000, @@ -55009,9 +56305,9 @@ "text" ], "cost": { - "input": 0.64, - "output": 2.588, - "cacheRead": 0, + "input": 1.1, + "output": 4.4, + "cacheRead": 0.55, "cacheWrite": 0 }, "contextWindow": 200000, @@ -55039,9 +56335,9 @@ "text" ], "cost": { - "input": 9.996, - "output": 19.992, - "cacheRead": 0, + "input": 1.1, + "output": 4.4, + "cacheRead": 0.55, "cacheWrite": 0 }, "contextWindow": 200000, @@ -55069,9 +56365,9 @@ "text" ], "cost": { - "input": 9.996, - "output": 19.992, - "cacheRead": 0, + "input": 22, + "output": 88, + "cacheRead": 11, "cacheWrite": 0 }, "contextWindow": 200000, @@ -55101,7 +56397,7 @@ "cost": { "input": 1.1, "output": 4.4, - "cacheRead": 0, + "cacheRead": 0.55, "cacheWrite": 0 }, "contextWindow": 200000, @@ -55164,7 +56460,7 @@ "cost": { "input": 1.1, "output": 4.4, - "cacheRead": 0, + "cacheRead": 0.55, "cacheWrite": 0 }, "contextWindow": 200000, @@ -55347,6 +56643,35 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "pokee-isaac": { + "id": "pokee-isaac", + "name": "Pokee-Isaac 28B", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.15, + "output": 1, + "cacheRead": 0.075, + "cacheWrite": 0 + }, + "contextWindow": 10000000, + "maxTokens": 60000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } + }, "poolside/laguna-m.1": { "id": "poolside/laguna-m.1", "name": "Laguna M.1", @@ -55378,7 +56703,7 @@ }, "poolside/laguna-s-2.1": { "id": "poolside/laguna-s-2.1", - "name": "poolside/laguna-s-2.1", + "name": "Laguna S 2.1", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -55387,9 +56712,9 @@ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.1, + "output": 0.2, + "cacheRead": 0.01, "cacheWrite": 0 }, "contextWindow": 1048576, @@ -55403,9 +56728,36 @@ "high", "xhigh" ] + } + }, + "poolside/laguna-s-2.1:thinking": { + "id": "poolside/laguna-s-2.1:thinking", + "name": "Laguna S 2.1 Thinking", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.1, + "output": 0.2, + "cacheRead": 0.01, + "cacheWrite": 0 }, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "contextWindow": 1048576, + "maxTokens": 131072, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "poolside/laguna-xs.2": { "id": "poolside/laguna-xs.2", @@ -55647,7 +56999,7 @@ "cost": { "input": 0.3, "output": 0.5, - "cacheRead": 0, + "cacheRead": 0.15, "cacheWrite": 0 }, "contextWindow": 41000, @@ -55696,7 +57048,7 @@ "cost": { "input": 0.13, "output": 0.5, - "cacheRead": 0, + "cacheRead": 0.065, "cacheWrite": 0 }, "contextWindow": 256000, @@ -55704,7 +57056,7 @@ }, "Qwen/Qwen3-235B-A22B-Instruct-2507": { "id": "Qwen/Qwen3-235B-A22B-Instruct-2507", - "name": "Qwen3 235B A22B-2507", + "name": "Qwen3 235B-A22B Instruct 2507", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -55736,7 +57088,7 @@ "cost": { "input": 0.13, "output": 0.5, - "cacheRead": 0, + "cacheRead": 0.065, "cacheWrite": 0 }, "contextWindow": 256000, @@ -55765,7 +57117,7 @@ }, "Qwen/Qwen3-235B-A22B-Thinking-2507": { "id": "Qwen/Qwen3-235B-A22B-Thinking-2507", - "name": "Qwen3 235B A22B Thinking-2507", + "name": "Qwen3-235B-A22B-Thinking-2507", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -55826,7 +57178,7 @@ }, "qwen/qwen3-32b": { "id": "qwen/qwen3-32b", - "name": "Qwen3-32B", + "name": "Qwen3 32B", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -55888,7 +57240,7 @@ "cost": { "input": 0.13, "output": 0.5, - "cacheRead": 0, + "cacheRead": 0.065, "cacheWrite": 0 }, "contextWindow": 262000, @@ -55926,9 +57278,9 @@ "text" ], "cost": { - "input": 0.15, + "input": 0.2, "output": 1.5, - "cacheRead": 0, + "cacheRead": 0.1, "cacheWrite": 0 }, "contextWindow": 262144, @@ -55989,7 +57341,7 @@ "cost": { "input": 0.15, "output": 0.65, - "cacheRead": 0, + "cacheRead": 0.075, "cacheWrite": 0 }, "contextWindow": 256000, @@ -56018,7 +57370,7 @@ }, "qwen/qwen3-next-80b-a3b-thinking": { "id": "qwen/qwen3-next-80b-a3b-thinking", - "name": "Qwen3 Next 80B A3B Thinking", + "name": "Qwen3-Next 80B-A3B (Thinking)", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -56070,7 +57422,7 @@ }, "qwen/qwen3.5-397b-a17b": { "id": "qwen/qwen3.5-397b-a17b", - "name": "Qwen3.5 397B A17B", + "name": "Qwen3.5 397B-A17B", "api": "openai-completions", "baseUrl": "https://nano-gpt.com/api/v1", "provider": "nanogpt", @@ -56108,7 +57460,7 @@ }, "qwen/qwen3.5-9b": { "id": "qwen/qwen3.5-9b", - "name": "Qwen3.5-9B", + "name": "Qwen3.5 9B", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -56303,7 +57655,7 @@ "cost": { "input": 0.1, "output": 0.4, - "cacheRead": 0, + "cacheRead": 0.05, "cacheWrite": 0 }, "contextWindow": 128000, @@ -56374,7 +57726,26 @@ }, "qwen3.5-122b-a10b": { "id": "qwen3.5-122b-a10b", - "name": "qwen3.5-122b-a10b", + "name": "Qwen3.5 122B A10B", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": false, + "input": [ + "text" + ], + "cost": { + "input": 0.437, + "output": 3.496, + "cacheRead": 0.103788, + "cacheWrite": 0 + }, + "contextWindow": 131072, + "maxTokens": 32768 + }, + "qwen3.5-122b-a10b:thinking": { + "id": "qwen3.5-122b-a10b:thinking", + "name": "Qwen3.5 122B A10B Thinking", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -56383,13 +57754,13 @@ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.437, + "output": 3.496, + "cacheRead": 0.103788, "cacheWrite": 0 }, - "contextWindow": 262144, - "maxTokens": 262144, + "contextWindow": 131072, + "maxTokens": 32768, "thinking": { "mode": "effort", "efforts": [ @@ -56398,9 +57769,7 @@ "medium", "high" ] - }, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + } }, "qwen3.5-27b": { "id": "qwen3.5-27b", @@ -57502,21 +58871,22 @@ }, "qwen3.7-flash": { "id": "qwen3.7-flash", - "name": "qwen3.7-flash", + "name": "Qwen3.7 Flash", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 0.03, + "output": 0.13, + "cacheRead": 0.006, + "cacheWrite": 0.038 }, - "contextWindow": 1000000, + "contextWindow": 991808, "maxTokens": 65536, "thinking": { "mode": "effort", @@ -57526,9 +58896,36 @@ "medium", "high" ] + } + }, + "qwen3.7-flash:thinking": { + "id": "qwen3.7-flash:thinking", + "name": "Qwen3.7 Flash Thinking", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.03, + "output": 0.13, + "cacheRead": 0.006, + "cacheWrite": 0.038 }, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "contextWindow": 983616, + "maxTokens": 65536, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + } }, "qwen3.7-max": { "id": "qwen3.7-max", @@ -57591,26 +58988,82 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, - "qwen3.8-max-preview": { - "id": "qwen3.8-max-preview", - "name": "qwen3.8-max-preview", + "qwen3.8-max": { + "id": "qwen3.8-max", + "name": "Qwen3.8 Max", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", "reasoning": false, + "input": [ + "text", + "image" + ], + "cost": { + "input": 2, + "output": 6, + "cacheRead": 0.25, + "cacheWrite": 2.5 + }, + "contextWindow": 991000, + "maxTokens": 65536 + }, + "qwen3.8-max-preview": { + "id": "qwen3.8-max-preview", + "name": "Qwen3.8 Max Preview", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": true, "input": [ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 + "input": 1.5, + "output": 5, + "cacheRead": 0.15, + "cacheWrite": 2 }, - "contextWindow": 983616, + "contextWindow": 991000, + "maxTokens": 64000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + } + }, + "qwen3.8-max:thinking": { + "id": "qwen3.8-max:thinking", + "name": "Qwen3.8 Max Thinking", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 2, + "output": 6, + "cacheRead": 0.25, + "cacheWrite": 2.5 + }, + "contextWindow": 991000, "maxTokens": 131072, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + } }, "qwq-32b": { "id": "qwq-32b", @@ -57719,45 +59172,63 @@ }, "sakana/fugu-ultra": { "id": "sakana/fugu-ultra", - "name": "sakana/fugu-ultra", + "name": "Fugu Ultra", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 5.25, + "output": 31.5, + "cacheRead": 0.525, "cacheWrite": 0 }, "contextWindow": 1000000, - "maxTokens": null, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "maxTokens": 16384, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "sakana/fugu-ultra-v1.1": { "id": "sakana/fugu-ultra-v1.1", - "name": "sakana/fugu-ultra-v1.1", + "name": "Fugu Ultra v1.1", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 5.25, + "output": 31.5, + "cacheRead": 0.525, "cacheWrite": 0 }, - "contextWindow": null, - "maxTokens": null, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "contextWindow": 1000000, + "maxTokens": 16384, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "Salesforce/Llama-xLAM-2-70b-fc-r": { "id": "Salesforce/Llama-xLAM-2-70b-fc-r", @@ -58450,7 +59921,7 @@ "cacheRead": 0.04, "cacheWrite": 0 }, - "contextWindow": 256000, + "contextWindow": 262144, "maxTokens": 256000, "thinking": { "mode": "effort", @@ -58655,7 +60126,7 @@ "cost": { "input": 0.15, "output": 0.7, - "cacheRead": 0, + "cacheRead": 0.075, "cacheWrite": 0 }, "contextWindow": 65536, @@ -58674,7 +60145,7 @@ "cost": { "input": 0.15, "output": 0.46, - "cacheRead": 0, + "cacheRead": 0.075, "cacheWrite": 0 }, "contextWindow": 262144, @@ -58885,7 +60356,7 @@ }, "TEE/glm-5.2": { "id": "TEE/glm-5.2", - "name": "TEE/glm-5.2", + "name": "GLM 5.2 TEE", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -58894,9 +60365,9 @@ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 1.4, + "output": 4.6, + "cacheRead": 0.5, "cacheWrite": 0 }, "contextWindow": 1048576, @@ -58910,9 +60381,36 @@ "high", "max" ] + } + }, + "TEE/glm-5.2:thinking": { + "id": "TEE/glm-5.2:thinking", + "name": "GLM 5.2 Thinking TEE", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 1.4, + "output": 4.6, + "cacheRead": 0.5, + "cacheWrite": 0 }, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "contextWindow": 1048576, + "maxTokens": 131072, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "max" + ] + } }, "TEE/gpt-oss-120b": { "id": "TEE/gpt-oss-120b", @@ -59038,24 +60536,36 @@ }, "TEE/kimi-k3": { "id": "TEE/kimi-k3", - "name": "TEE/kimi-k3", + "name": "Kimi K3 TEE", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": false, + "reasoning": true, "input": [ - "text" + "text", + "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 3, + "output": 15, + "cacheRead": 1.5, "cacheWrite": 0 }, "contextWindow": 1048576, - "maxTokens": 131072, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "maxTokens": 65535, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "high", + "max" + ], + "defaultLevel": "max", + "effortMap": { + "max": "max" + }, + "requiresEffort": true + } }, "TEE/llama3-3-70b": { "id": "TEE/llama3-3-70b", @@ -59133,7 +60643,7 @@ "cost": { "input": 0.2, "output": 1.38, - "cacheRead": 0, + "cacheRead": 0.1, "cacheWrite": 0 }, "contextWindow": 196608, @@ -59245,7 +60755,7 @@ "cost": { "input": 0.46, "output": 3.68, - "cacheRead": 0, + "cacheRead": 0.23, "cacheWrite": 0 }, "contextWindow": 262144, @@ -59358,7 +60868,7 @@ "cost": { "input": 0.3, "output": 1.5, - "cacheRead": 0, + "cacheRead": 0.15, "cacheWrite": 0 }, "contextWindow": 131072, @@ -59396,28 +60906,36 @@ }, "tencent/hy3": { "id": "tencent/hy3", - "name": "tencent/hy3", + "name": "Tencent Hy3", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": false, + "reasoning": true, "input": [ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.066, + "output": 0.26, + "cacheRead": 0.029, "cacheWrite": 0 }, "contextWindow": 262144, "maxTokens": 262144, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "tencent/hy3-preview": { "id": "tencent/hy3-preview", - "name": "Hy3 Preview", + "name": "Hy3 preview", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -59683,6 +61201,76 @@ "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": false, + "input": [ + "text", + "image" + ], + "cost": { + "input": 1, + "output": 4.05, + "cacheRead": 0.17, + "cacheWrite": 0 + }, + "contextWindow": 1048000, + "maxTokens": 32768 + }, + "thinkingmachines/Inkling-Small": { + "id": "thinkingmachines/Inkling-Small", + "name": "Inkling Small", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": false, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.5, + "output": 1.2, + "cacheRead": 0.1, + "cacheWrite": 0 + }, + "contextWindow": 524288, + "maxTokens": 32768 + }, + "thinkingmachines/Inkling-Small:thinking": { + "id": "thinkingmachines/Inkling-Small:thinking", + "name": "Inkling Small Thinking", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.5, + "output": 1.2, + "cacheRead": 0.1, + "cacheWrite": 0 + }, + "contextWindow": 524288, + "maxTokens": 32768, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } + }, + "thinkingmachines/inkling:thinking": { + "id": "thinkingmachines/inkling:thinking", + "name": "Inkling Thinking", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", "reasoning": true, "input": [ "text", @@ -59691,10 +61279,10 @@ "cost": { "input": 1, "output": 4.05, - "cacheRead": 0.16999999999999998, + "cacheRead": 0.17, "cacheWrite": 0 }, - "contextWindow": 1048576, + "contextWindow": 1048000, "maxTokens": 32768, "thinking": { "mode": "effort", @@ -59705,40 +61293,7 @@ "high", "xhigh" ] - }, - "supportsComputerUse": false, - "supportsComputerUseConfig": false - }, - "thinkingmachines/Inkling-Small": { - "id": "thinkingmachines/Inkling-Small", - "name": "thinkingmachines/Inkling-Small", - "api": "openai-completions", - "provider": "nanogpt", - "baseUrl": "https://nano-gpt.com/api/v1", - "reasoning": true, - "input": [ - "text" - ], - "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, - "cacheWrite": 0 - }, - "contextWindow": 1048576, - "maxTokens": 32768, - "thinking": { - "mode": "effort", - "efforts": [ - "minimal", - "low", - "medium", - "high", - "xhigh" - ] - }, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + } }, "THUDM/GLM-4-32B-0414": { "id": "THUDM/GLM-4-32B-0414", @@ -60332,7 +61887,7 @@ "cost": { "input": 2, "output": 6, - "cacheRead": 0, + "cacheRead": 1, "cacheWrite": 0 }, "contextWindow": 2000000, @@ -60483,9 +62038,9 @@ "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 2, + "output": 6, + "cacheRead": 0.5, "cacheWrite": 0 }, "contextWindow": 500000, @@ -60499,9 +62054,7 @@ "high", "xhigh" ] - }, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + } }, "x-ai/grok-build-0.1": { "id": "x-ai/grok-build-0.1", @@ -60576,13 +62129,13 @@ "image" ], "cost": { - "input": 1.25, - "output": 2.5, - "cacheRead": 0.2, + "input": 2, + "output": 6, + "cacheRead": 0.5, "cacheWrite": 0 }, - "contextWindow": 1000000, - "maxTokens": 1000000, + "contextWindow": 500000, + "maxTokens": 500000, "thinking": { "mode": "effort", "efforts": [ @@ -60786,7 +62339,7 @@ }, "xiaomi/mimo-v2.5-pro-crof": { "id": "xiaomi/mimo-v2.5-pro-crof", - "name": "xiaomi/mimo-v2.5-pro-crof", + "name": "MiMo V2.5 Pro (Crof)", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -60795,13 +62348,13 @@ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.4, + "output": 0.8, + "cacheRead": 0.003, "cacheWrite": 0 }, - "contextWindow": null, - "maxTokens": null, + "contextWindow": 1000000, + "maxTokens": 131072, "thinking": { "mode": "effort", "efforts": [ @@ -60809,9 +62362,34 @@ "medium", "high" ] + } + }, + "xiaomi/mimo-v2.5-pro-crof:thinking": { + "id": "xiaomi/mimo-v2.5-pro-crof:thinking", + "name": "MiMo V2.5 Pro Thinking (Crof)", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.4, + "output": 0.8, + "cacheRead": 0.003, + "cacheWrite": 0 }, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "contextWindow": 1000000, + "maxTokens": 131072, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "high" + ] + } }, "xiaomi/mimo-v2.5-pro-ultraspeed": { "id": "xiaomi/mimo-v2.5-pro-ultraspeed", @@ -60834,6 +62412,61 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "xiaomi/mimo-v2.5-pro:thinking": { + "id": "xiaomi/mimo-v2.5-pro:thinking", + "name": "MiMo V2.5 Pro Thinking", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.435, + "output": 0.87, + "cacheRead": 0.0036, + "cacheWrite": 0 + }, + "contextWindow": 1048576, + "maxTokens": 131072, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "high" + ] + } + }, + "xiaomi/mimo-v2.5:thinking": { + "id": "xiaomi/mimo-v2.5:thinking", + "name": "MiMo V2.5 Thinking", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.14, + "output": 0.28, + "cacheRead": 0.0028, + "cacheWrite": 0 + }, + "contextWindow": 1048576, + "maxTokens": 131072, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "high" + ] + } + }, "yi-large": { "id": "yi-large", "name": "yi-large", @@ -60899,7 +62532,7 @@ }, "z-ai/glm-4.5v": { "id": "z-ai/glm-4.5v", - "name": "GLM 4.5V", + "name": "GLM-4.5V", "api": "openai-completions", "baseUrl": "https://nano-gpt.com/api/v1", "provider": "nanogpt", @@ -60940,9 +62573,9 @@ "text" ], "cost": { - "input": 0.4, - "output": 1.5, - "cacheRead": 0, + "input": 0.35, + "output": 1.4, + "cacheRead": 0.175, "cacheWrite": 0 }, "contextWindow": 200000, @@ -60969,9 +62602,9 @@ "text" ], "cost": { - "input": 0.4, - "output": 1.5, - "cacheRead": 0, + "input": 0.35, + "output": 1.4, + "cacheRead": 0.175, "cacheWrite": 0 }, "contextWindow": 200000, @@ -61024,7 +62657,7 @@ "cacheWrite": 0 }, "contextWindow": 202800, - "maxTokens": 131100 + "maxTokens": 131072 }, "z-ai/glm-5v-turbo:thinking": { "id": "z-ai/glm-5v-turbo:thinking", @@ -61044,7 +62677,7 @@ "cacheWrite": 0 }, "contextWindow": 202800, - "maxTokens": 131100, + "maxTokens": 131072, "thinking": { "mode": "effort", "efforts": [ @@ -61090,7 +62723,7 @@ "cost": { "input": 0.12, "output": 0.8, - "cacheRead": 0, + "cacheRead": 0.06, "cacheWrite": 0 }, "contextWindow": 128000, @@ -61109,7 +62742,7 @@ "cost": { "input": 0.12, "output": 0.8, - "cacheRead": 0, + "cacheRead": 0.06, "cacheWrite": 0 }, "contextWindow": 128000, @@ -61251,13 +62884,13 @@ "text" ], "cost": { - "input": 0.15, + "input": 0.2, "output": 0.8, - "cacheRead": 0, + "cacheRead": 0.1, "cacheWrite": 0 }, "contextWindow": 200000, - "maxTokens": 128000, + "maxTokens": 65535, "thinking": { "mode": "effort", "efforts": [ @@ -61282,7 +62915,7 @@ "cost": { "input": 0.07, "output": 0.4, - "cacheRead": 0, + "cacheRead": 0.035, "cacheWrite": 0 }, "contextWindow": 200000, @@ -61311,7 +62944,7 @@ "cost": { "input": 0.07, "output": 0.4, - "cacheRead": 0, + "cacheRead": 0.035, "cacheWrite": 0 }, "contextWindow": 200000, @@ -61398,7 +63031,7 @@ "cost": { "input": 0.2, "output": 0.8, - "cacheRead": 0, + "cacheRead": 0.1, "cacheWrite": 0 }, "contextWindow": 200000, @@ -61425,9 +63058,9 @@ "text" ], "cost": { - "input": 0.3, + "input": 0.5, "output": 2.55, - "cacheRead": 0, + "cacheRead": 0.13, "cacheWrite": 0 }, "contextWindow": 200000, @@ -61533,9 +63166,9 @@ "text" ], "cost": { - "input": 0.3, + "input": 0.5, "output": 2.55, - "cacheRead": 0, + "cacheRead": 0.13, "cacheWrite": 0 }, "contextWindow": 200000, @@ -61562,9 +63195,9 @@ "text" ], "cost": { - "input": 0.3, - "output": 2.55, - "cacheRead": 0, + "input": 0.75, + "output": 2.6, + "cacheRead": 0.15, "cacheWrite": 0 }, "contextWindow": 200000, @@ -61591,9 +63224,9 @@ "text" ], "cost": { - "input": 0.3, - "output": 2.55, - "cacheRead": 0, + "input": 0.75, + "output": 2.6, + "cacheRead": 0.15, "cacheWrite": 0 }, "contextWindow": 200000, @@ -61611,7 +63244,7 @@ }, "zai-org/glm-5.2": { "id": "zai-org/glm-5.2", - "name": "zai-org/glm-5.2", + "name": "GLM 5.2", "api": "openai-completions", "provider": "nanogpt", "baseUrl": "https://nano-gpt.com/api/v1", @@ -61620,9 +63253,9 @@ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.42, + "output": 1.32, + "cacheRead": 0.078, "cacheWrite": 0 }, "contextWindow": 1048576, @@ -61636,9 +63269,36 @@ "high", "max" ] + } + }, + "zai-org/glm-5.2:thinking": { + "id": "zai-org/glm-5.2:thinking", + "name": "GLM 5.2 Thinking", + "api": "openai-completions", + "provider": "nanogpt", + "baseUrl": "https://nano-gpt.com/api/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.42, + "output": 1.32, + "cacheRead": 0.078, + "cacheWrite": 0 }, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "contextWindow": 1048576, + "maxTokens": 131072, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "max" + ] + } }, "zai-org/glm-latest": { "id": "zai-org/glm-latest", @@ -61651,12 +63311,12 @@ "text" ], "cost": { - "input": 0.75, - "output": 2.6, - "cacheRead": 0.15, + "input": 0.42, + "output": 1.32, + "cacheRead": 0.078, "cacheWrite": 0 }, - "contextWindow": 200000, + "contextWindow": 1048576, "maxTokens": 131072, "thinking": { "mode": "effort", @@ -61872,7 +63532,7 @@ }, "deepseek/deepseek-r1": { "id": "deepseek/deepseek-r1", - "name": "R1", + "name": "DeepSeek-R1", "api": "openai-completions", "provider": "novita", "baseUrl": "https://api.novita.ai/openai/v1", @@ -62241,6 +63901,7 @@ "thinking": { "mode": "effort", "efforts": [ + "low", "high", "max" ] @@ -62250,7 +63911,7 @@ }, "deepseek/deepseek-v4-flash-0731": { "id": "deepseek/deepseek-v4-flash-0731", - "name": "Deepseek V4 Flash 0731", + "name": "DeepSeek V4 Flash 0731", "api": "openai-completions", "provider": "novita", "baseUrl": "https://api.novita.ai/openai/v1", @@ -62270,6 +63931,7 @@ "thinking": { "mode": "effort", "efforts": [ + "low", "high", "max" ] @@ -62376,7 +64038,7 @@ }, "google/gemma-4-26b-a4b-it": { "id": "google/gemma-4-26b-a4b-it", - "name": "Gemma 4 26B A4B", + "name": "Gemma 4 26B A4B IT", "api": "openai-completions", "provider": "novita", "baseUrl": "https://api.novita.ai/openai/v1", @@ -62409,7 +64071,7 @@ }, "google/gemma-4-31b-it": { "id": "google/gemma-4-31b-it", - "name": "Gemma 4 31B", + "name": "Gemma 4 31B IT", "api": "openai-completions", "provider": "novita", "baseUrl": "https://api.novita.ai/openai/v1", @@ -62509,7 +64171,7 @@ }, "inclusionai/ling-2.6-flash": { "id": "inclusionai/ling-2.6-flash", - "name": "Ling-2.6 Flash", + "name": "Ling-2.6-flash", "api": "openai-completions", "provider": "novita", "baseUrl": "https://api.novita.ai/openai/v1", @@ -62539,6 +64201,38 @@ "input": [ "text" ], + "cost": { + "input": 0.06, + "output": 0.18, + "cacheRead": 0.012, + "cacheWrite": 0 + }, + "contextWindow": 262144, + "maxTokens": 32768, + "supportsTools": true, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + }, + "supportsComputerUse": false, + "supportsComputerUseConfig": false + }, + "inclusionai/ling-3.0-tiny": { + "id": "inclusionai/ling-3.0-tiny", + "name": "Ling 3.0 Tiny", + "api": "openai-completions", + "provider": "novita", + "baseUrl": "https://api.novita.ai/openai/v1", + "reasoning": true, + "input": [ + "text" + ], "cost": { "input": 0, "output": 0, @@ -62558,8 +64252,7 @@ "xhigh" ] }, - "supportsComputerUse": false, - "supportsComputerUseConfig": false + "supportsComputerUse": false }, "inclusionai/ring-2.6-1t": { "id": "inclusionai/ring-2.6-1t", @@ -62659,31 +64352,9 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, - "meta-llama/llama-3.2-3b-instruct": { - "id": "meta-llama/llama-3.2-3b-instruct", - "name": "Llama 3.2 3B Instruct", - "api": "openai-completions", - "provider": "novita", - "baseUrl": "https://api.novita.ai/openai/v1", - "reasoning": false, - "input": [ - "text" - ], - "cost": { - "input": 0.03, - "output": 0.05, - "cacheRead": 0, - "cacheWrite": 0 - }, - "contextWindow": 32768, - "maxTokens": 32000, - "supportsTools": false, - "supportsComputerUse": false, - "supportsComputerUseConfig": false - }, "meta-llama/llama-3.3-70b-instruct": { "id": "meta-llama/llama-3.3-70b-instruct", - "name": "Llama 3.3 70B Instruct", + "name": "Llama-3.3-70B-Instruct", "api": "openai-completions", "provider": "novita", "baseUrl": "https://api.novita.ai/openai/v1", @@ -62697,8 +64368,8 @@ "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 6000, - "maxTokens": 120000, + "contextWindow": 12288, + "maxTokens": 12288, "supportsTools": true, "supportsComputerUse": false, "supportsComputerUseConfig": false @@ -62728,7 +64399,7 @@ }, "meta-llama/llama-4-scout-17b-16e-instruct": { "id": "meta-llama/llama-4-scout-17b-16e-instruct", - "name": "Llama 4 Scout 17B 16E", + "name": "Llama 4 Scout Instruct", "api": "openai-completions", "provider": "novita", "baseUrl": "https://api.novita.ai/openai/v1", @@ -62814,9 +64485,9 @@ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 1.5, + "output": 4.5, + "cacheRead": 0.3, "cacheWrite": 0 }, "contextWindow": 1048576, @@ -62837,7 +64508,7 @@ }, "minimax/minimax-m2": { "id": "minimax/minimax-m2", - "name": "MiniMax M2", + "name": "MiniMax-M2", "api": "openai-completions", "provider": "novita", "baseUrl": "https://api.novita.ai/openai/v1", @@ -62868,7 +64539,7 @@ }, "minimax/minimax-m2.1": { "id": "minimax/minimax-m2.1", - "name": "MiniMax M2.1", + "name": "MiniMax-M2.1", "api": "openai-completions", "provider": "novita", "baseUrl": "https://api.novita.ai/openai/v1", @@ -62961,7 +64632,7 @@ }, "minimax/minimax-m2.7": { "id": "minimax/minimax-m2.7", - "name": "MiniMax M2.7", + "name": "MiniMax-M2.7", "api": "openai-completions", "provider": "novita", "baseUrl": "https://api.novita.ai/openai/v1", @@ -63023,7 +64694,7 @@ }, "minimax/minimax-m3": { "id": "minimax/minimax-m3", - "name": "MiniMax M3", + "name": "MiniMax-M3", "api": "openai-completions", "provider": "novita", "baseUrl": "https://api.novita.ai/openai/v1", @@ -63588,7 +65259,7 @@ }, "qwen/qwen3-coder-30b-a3b-instruct": { "id": "qwen/qwen3-coder-30b-a3b-instruct", - "name": "Qwen3 Coder 30B A3B Instruct", + "name": "Qwen3-Coder 30B-A3B Instruct", "api": "openai-completions", "provider": "novita", "baseUrl": "https://api.novita.ai/openai/v1", @@ -63865,7 +65536,7 @@ }, "qwen/qwen3.5-27b": { "id": "qwen/qwen3.5-27b", - "name": "Qwen3.5-27B", + "name": "Qwen3.5 27B", "api": "openai-completions", "provider": "novita", "baseUrl": "https://api.novita.ai/openai/v1", @@ -63897,7 +65568,7 @@ }, "qwen/qwen3.5-35b-a3b": { "id": "qwen/qwen3.5-35b-a3b", - "name": "Qwen3.5-35B-A3B", + "name": "Qwen3.5 35B-A3B", "api": "openai-completions", "provider": "novita", "baseUrl": "https://api.novita.ai/openai/v1", @@ -63929,7 +65600,7 @@ }, "qwen/qwen3.5-397b-a17b": { "id": "qwen/qwen3.5-397b-a17b", - "name": "Qwen3.5 397B A17B", + "name": "Qwen3.5 397B-A17B", "api": "openai-completions", "provider": "novita", "baseUrl": "https://api.novita.ai/openai/v1", @@ -64025,7 +65696,7 @@ }, "qwen/qwen3.6-35b-a3b": { "id": "qwen/qwen3.6-35b-a3b", - "name": "Qwen3.6-35B-A3B", + "name": "Qwen3.6 35B-A3B", "api": "openai-completions", "provider": "novita", "baseUrl": "https://api.novita.ai/openai/v1", @@ -64118,6 +65789,37 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "qwen/qwen3.8-max": { + "id": "qwen/qwen3.8-max", + "name": "Qwen3.8 Max", + "api": "openai-completions", + "provider": "novita", + "baseUrl": "https://api.novita.ai/openai/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 2, + "output": 6, + "cacheRead": 0.25, + "cacheWrite": 0 + }, + "contextWindow": 1000000, + "maxTokens": 131072, + "supportsTools": true, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + }, + "supportsComputerUse": false + }, "sao10k/l3-70b-euryale-v2.1": { "id": "sao10k/l3-70b-euryale-v2.1", "name": "L3 70B Euryale V2.1", @@ -64251,7 +65953,7 @@ }, "tencent/hy3": { "id": "tencent/hy3", - "name": "Hy3", + "name": "Tencent Hy3", "api": "openai-completions", "provider": "novita", "baseUrl": "https://api.novita.ai/openai/v1", @@ -65110,6 +66812,7 @@ "thinking": { "mode": "effort", "efforts": [ + "low", "high", "max" ] @@ -67792,7 +69495,7 @@ }, "qwen/qwen3-235b-a22b": { "id": "qwen/qwen3-235b-a22b", - "name": "Qwen3 235B A22B", + "name": "Qwen3 235B-A22B", "api": "openai-completions", "provider": "nvidia", "baseUrl": "https://integrate.api.nvidia.com/v1", @@ -67858,7 +69561,7 @@ }, "qwen/qwen3-next-80b-a3b-thinking": { "id": "qwen/qwen3-next-80b-a3b-thinking", - "name": "Qwen3 Next 80B A3B Thinking", + "name": "Qwen3-Next 80B-A3B (Thinking)", "api": "openai-completions", "provider": "nvidia", "baseUrl": "https://integrate.api.nvidia.com/v1", @@ -68205,7 +69908,7 @@ }, "z-ai/glm-5.1": { "id": "z-ai/glm-5.1", - "name": "GLM 5.1", + "name": "GLM-5.1", "api": "openai-completions", "provider": "nvidia", "baseUrl": "https://integrate.api.nvidia.com/v1", @@ -68465,21 +70168,11 @@ }, "deepseek-v4-flash:0731": { "id": "deepseek-v4-flash:0731", - "name": "deepseek-v4-flash:0731", + "name": "DeepSeek V4 Flash 0731", "api": "ollama-chat", "provider": "ollama-cloud", "baseUrl": "https://ollama.com", "reasoning": true, - "thinking": { - "mode": "effort", - "efforts": [ - "minimal", - "low", - "medium", - "high", - "xhigh" - ] - }, "input": [ "text" ], @@ -68492,7 +70185,15 @@ "contextWindow": 1048576, "maxTokens": 65536, "omitMaxOutputTokens": true, - "supportsComputerUse": false + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + } }, "deepseek-v4-pro": { "id": "deepseek-v4-pro", @@ -71448,7 +73149,7 @@ "opencode-go": { "deepseek-v4-flash": { "id": "deepseek-v4-flash", - "name": "DeepSeek V4 Flash (New)", + "name": "DeepSeek V4 Flash (2x usage)", "api": "openai-completions", "provider": "opencode-go", "baseUrl": "https://opencode.ai/zen/go/v1", @@ -71457,9 +73158,9 @@ "text" ], "cost": { - "input": 0.14, - "output": 0.28, - "cacheRead": 0.0028, + "input": 0.07, + "output": 0.14, + "cacheRead": 0.0014, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -71473,6 +73174,7 @@ "thinking": { "mode": "effort", "efforts": [ + "low", "high", "max" ] @@ -72141,6 +73843,36 @@ "xhigh" ] } + }, + "qwen3.8-max": { + "id": "qwen3.8-max", + "name": "Qwen3.8 Max", + "api": "anthropic-messages", + "provider": "opencode-go", + "baseUrl": "https://opencode.ai/zen/go", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 2, + "output": 6, + "cacheRead": 0.25, + "cacheWrite": 2.5 + }, + "contextWindow": 1000000, + "maxTokens": 131072, + "thinking": { + "mode": "budget", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } } }, "opencode-zen": { @@ -72573,6 +74305,7 @@ "thinking": { "mode": "effort", "efforts": [ + "low", "high", "max" ] @@ -72580,7 +74313,7 @@ }, "deepseek-v4-flash-free": { "id": "deepseek-v4-flash-free", - "name": "DeepSeek V4 Flash Free (New)", + "name": "DeepSeek V4 Flash Free", "api": "openai-completions", "provider": "opencode-zen", "baseUrl": "https://opencode.ai/zen/v1", @@ -72599,6 +74332,7 @@ "thinking": { "mode": "effort", "efforts": [ + "low", "high", "max" ] @@ -73567,7 +75301,7 @@ "grok-build-0.1": { "id": "grok-build-0.1", "name": "Grok Build 0.1", - "api": "openai-completions", + "api": "openai-responses", "provider": "opencode-zen", "baseUrl": "https://opencode.ai/zen/v1", "reasoning": true, @@ -73889,6 +75623,64 @@ ] } }, + "ling-3.0-tiny-free": { + "id": "ling-3.0-tiny-free", + "name": "Ling-3.0-tiny Free", + "api": "openai-completions", + "provider": "opencode-zen", + "baseUrl": "https://opencode.ai/zen/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0, + "output": 0, + "cacheRead": 0, + "cacheWrite": 0 + }, + "contextWindow": 262144, + "maxTokens": 32768, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } + }, + "longcat-2.0-free": { + "id": "longcat-2.0-free", + "name": "LongCat-2.0 Free", + "api": "openai-completions", + "provider": "opencode-zen", + "baseUrl": "https://opencode.ai/zen/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0, + "output": 0, + "cacheRead": 0, + "cacheWrite": 0 + }, + "contextWindow": 1000000, + "maxTokens": 131072, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } + }, "mimo-v2-flash-free": { "id": "mimo-v2-flash-free", "name": "MiMo V2 Flash Free", @@ -74434,7 +76226,7 @@ }, "~anthropic/claude-haiku-latest": { "id": "~anthropic/claude-haiku-latest", - "name": "Claude Haiku Latest", + "name": "Anthropic Claude Haiku Latest", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -74496,7 +76288,7 @@ }, "~anthropic/claude-sonnet-latest": { "id": "~anthropic/claude-sonnet-latest", - "name": "Claude Sonnet Latest", + "name": "Anthropic Claude Sonnet Latest", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -74542,11 +76334,13 @@ "cacheWrite": 0 }, "contextWindow": 1048576, - "maxTokens": 65536, + "maxTokens": 384000, "thinking": { "mode": "effort", "efforts": [ - "high" + "low", + "high", + "max" ] }, "supportsComputerUse": false, @@ -74554,7 +76348,7 @@ }, "~google/gemini-flash-latest": { "id": "~google/gemini-flash-latest", - "name": "Gemini Flash Latest", + "name": "Google Gemini Flash Latest", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -74567,7 +76361,7 @@ "input": 1.5, "output": 7.5, "cacheRead": 0.15, - "cacheWrite": 0.08333333333333334 + "cacheWrite": 0.0833333333333333 }, "contextWindow": 1048576, "maxTokens": 65536, @@ -74585,7 +76379,7 @@ }, "~google/gemini-pro-latest": { "id": "~google/gemini-pro-latest", - "name": "Gemini Pro Latest", + "name": "Google Gemini Pro Latest", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -74616,7 +76410,7 @@ }, "~moonshotai/kimi-latest": { "id": "~moonshotai/kimi-latest", - "name": "Kimi Latest", + "name": "MoonshotAI Kimi Latest", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -74626,7 +76420,7 @@ "image" ], "cost": { - "input": 2.9000000000000004, + "input": 2.5, "output": 14, "cacheRead": 0.29, "cacheWrite": 0 @@ -74647,7 +76441,7 @@ }, "~openai/gpt-latest": { "id": "~openai/gpt-latest", - "name": "GPT Latest", + "name": "OpenAI GPT Latest", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -74678,7 +76472,7 @@ }, "~openai/gpt-mini-latest": { "id": "~openai/gpt-mini-latest", - "name": "GPT Mini Latest", + "name": "OpenAI GPT Mini Latest", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -74725,7 +76519,7 @@ "cacheWrite": 0 }, "contextWindow": 500000, - "maxTokens": null, + "maxTokens": 500000, "thinking": { "mode": "effort", "efforts": [ @@ -75566,7 +77360,7 @@ }, "anthropic/claude-opus-4.7-fast": { "id": "anthropic/claude-opus-4.7-fast", - "name": "Claude Opus 4.7 (Fast)", + "name": "Claude Opus 4.7", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -75662,7 +77456,7 @@ }, "anthropic/claude-opus-4.8-fast": { "id": "anthropic/claude-opus-4.8-fast", - "name": "Claude Opus 4.8 (Fast)", + "name": "Claude Opus 4.8", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -75758,7 +77552,7 @@ }, "anthropic/claude-opus-5-fast": { "id": "anthropic/claude-opus-5-fast", - "name": "Claude Opus 5 (Fast)", + "name": "Claude Opus 5", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -75788,6 +77582,37 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "anthropic/claude-opus-5:batch": { + "id": "anthropic/claude-opus-5:batch", + "name": "Claude Opus 5 (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 2.5, + "output": 12.5, + "cacheRead": 0.25, + "cacheWrite": 3.125 + }, + "contextWindow": 1000000, + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "high", + "xhigh", + "max" + ] + }, + "supportsComputerUse": false + }, "anthropic/claude-sonnet-4": { "id": "anthropic/claude-sonnet-4", "name": "Claude Sonnet 4", @@ -75912,6 +77737,36 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "anthropic/claude-sonnet-4.6:batch": { + "id": "anthropic/claude-sonnet-4.6:batch", + "name": "Claude Sonnet 4.6 (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 1.5, + "output": 7.5, + "cacheRead": 0.15, + "cacheWrite": 1.875 + }, + "contextWindow": 1000000, + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + }, + "supportsComputerUse": false + }, "anthropic/claude-sonnet-5": { "id": "anthropic/claude-sonnet-5", "name": "Claude Sonnet 5", @@ -76409,7 +78264,7 @@ }, "cohere/command-r-08-2024": { "id": "cohere/command-r-08-2024", - "name": "Command R (08-2024)", + "name": "Command R", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -76481,7 +78336,7 @@ }, "deepseek/deepseek-chat": { "id": "deepseek/deepseek-chat", - "name": "DeepSeek V3", + "name": "DeepSeek-V3.2 (Non-thinking Mode)", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -76550,7 +78405,7 @@ }, "deepseek/deepseek-r1": { "id": "deepseek/deepseek-r1", - "name": "R1", + "name": "DeepSeek-R1", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -76667,13 +78522,13 @@ "text" ], "cost": { - "input": 0.26899999999999996, - "output": 0.39999999999999997, - "cacheRead": 0.13449999999999998, + "input": 0.26, + "output": 0.38, + "cacheRead": 0.13, "cacheWrite": 0 }, "contextWindow": 163840, - "maxTokens": 65536, + "maxTokens": 163840, "thinking": { "mode": "effort", "efforts": [ @@ -76731,7 +78586,9 @@ "thinking": { "mode": "effort", "efforts": [ - "high" + "low", + "high", + "max" ] }, "supportsComputerUse": false, @@ -76754,7 +78611,7 @@ "cacheWrite": 0 }, "contextWindow": 1048576, - "maxTokens": 65536, + "maxTokens": 384000, "thinking": { "mode": "effort", "efforts": [ @@ -76787,7 +78644,9 @@ "thinking": { "mode": "effort", "efforts": [ - "high" + "low", + "high", + "max" ] }, "supportsComputerUse": false, @@ -76900,7 +78759,7 @@ "input": 0.3, "output": 2.5, "cacheRead": 0.03, - "cacheWrite": 0.08333333333333334 + "cacheWrite": 0.0833333333333333 }, "contextWindow": 1048576, "maxTokens": 65535, @@ -76918,7 +78777,7 @@ }, "google/gemini-2.5-flash-lite": { "id": "google/gemini-2.5-flash-lite", - "name": "Gemini 2.5 Flash Lite", + "name": "Gemini 2.5 Flash-Lite", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -76931,7 +78790,7 @@ "input": 0.09999999999999999, "output": 0.39999999999999997, "cacheRead": 0.01, - "cacheWrite": 0.08333333333333334 + "cacheWrite": 0.0833333333333333 }, "contextWindow": 1048576, "maxTokens": 65535, @@ -77214,10 +79073,10 @@ "input": 0.5, "output": 3, "cacheRead": 0.049999999999999996, - "cacheWrite": 0.08333333333333334 + "cacheWrite": 0.0833333333333333 }, "contextWindow": 1048576, - "maxTokens": 65535, + "maxTokens": 65536, "thinking": { "mode": "effort", "efforts": [ @@ -77249,7 +79108,7 @@ "cacheWrite": 0 }, "contextWindow": 1048576, - "maxTokens": 65535, + "maxTokens": 65536, "thinking": { "mode": "effort", "efforts": [ @@ -77265,7 +79124,7 @@ }, "google/gemini-3-pro-image": { "id": "google/gemini-3-pro-image", - "name": "Nano Banana Pro (Gemini 3 Pro Image)", + "name": "Nano Banana Pro", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -77338,7 +79197,7 @@ "input": 0.25, "output": 1.5, "cacheRead": 0.024999999999999998, - "cacheWrite": 0.08333333333333334 + "cacheWrite": 0.0833333333333333 }, "contextWindow": 1048576, "maxTokens": 65536, @@ -77370,7 +79229,7 @@ "input": 0.25, "output": 1.5, "cacheRead": 0.024999999999999998, - "cacheWrite": 0.08333333333333334 + "cacheWrite": 0.0833333333333333 }, "contextWindow": 1048576, "maxTokens": 65536, @@ -77514,7 +79373,7 @@ "input": 1.5, "output": 9, "cacheRead": 0.15, - "cacheWrite": 0.08333333333333334 + "cacheWrite": 0.0833333333333333 }, "contextWindow": 1048576, "maxTokens": 65536, @@ -77546,7 +79405,7 @@ "input": 0.3, "output": 2.5, "cacheRead": 0.03, - "cacheWrite": 0.08333333333333334 + "cacheWrite": 0.0833333333333333 }, "contextWindow": 1048576, "maxTokens": 65536, @@ -77642,7 +79501,7 @@ "input": 1.5, "output": 7.5, "cacheRead": 0.15, - "cacheWrite": 0.08333333333333334 + "cacheWrite": 0.0833333333333333 }, "contextWindow": 1048576, "maxTokens": 65536, @@ -77674,7 +79533,7 @@ "input": 0.75, "output": 3.75, "cacheRead": 0.075, - "cacheWrite": 0.08333333333333334 + "cacheWrite": 0.0833333333333333 }, "contextWindow": 1048576, "maxTokens": 65536, @@ -77759,7 +79618,7 @@ }, "google/gemma-4-26b-a4b-it": { "id": "google/gemma-4-26b-a4b-it", - "name": "Gemma 4 26B A4B", + "name": "Gemma 4 26B A4B IT", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -77821,7 +79680,7 @@ }, "google/gemma-4-31b-it": { "id": "google/gemma-4-31b-it", - "name": "Gemma 4 31B", + "name": "Gemma 4 31B IT", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -78018,7 +79877,7 @@ }, "inclusionai/ling-2.6-flash": { "id": "inclusionai/ling-2.6-flash", - "name": "Ling-2.6 Flash", + "name": "Ling-2.6-flash", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -78058,6 +79917,35 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "inclusionai/ling-3.0-flash": { + "id": "inclusionai/ling-3.0-flash", + "name": "Ling-3.0-flash", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.020999999999999998, + "output": 0.063, + "cacheRead": 0.004200000000000001, + "cacheWrite": 0 + }, + "contextWindow": 262144, + "maxTokens": 32768, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + }, + "supportsComputerUse": false + }, "inclusionai/ling-3.0-flash:free": { "id": "inclusionai/ling-3.0-flash:free", "name": "Ling-3.0-flash (free)", @@ -78088,6 +79976,35 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "inclusionai/ling-3.0-tiny:free": { + "id": "inclusionai/ling-3.0-tiny:free", + "name": "Ling 3.0 Tiny (free)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0, + "output": 0, + "cacheRead": 0, + "cacheWrite": 0 + }, + "contextWindow": 262144, + "maxTokens": 32768, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + }, + "supportsComputerUse": false + }, "inclusionai/ring-2.6-1t": { "id": "inclusionai/ring-2.6-1t", "name": "Ring-2.6-1T", @@ -78400,7 +80317,7 @@ }, "meta-llama/llama-3.3-70b-instruct": { "id": "meta-llama/llama-3.3-70b-instruct", - "name": "Llama 3.3 70B Instruct", + "name": "Llama-3.3-70B-Instruct", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -78409,13 +80326,13 @@ "text" ], "cost": { - "input": 0.13, - "output": 0.39999999999999997, + "input": 0.09999999999999999, + "output": 0.32, "cacheRead": 0, "cacheWrite": 0 }, "contextWindow": 131072, - "maxTokens": 128000, + "maxTokens": 16384, "supportsComputerUse": false, "supportsComputerUseConfig": false }, @@ -78515,6 +80432,36 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "meta/muse-spark-1.2": { + "id": "meta/muse-spark-1.2", + "name": "Muse Spark 1.2", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 1.25, + "output": 4.25, + "cacheRead": 0.15, + "cacheWrite": 0 + }, + "contextWindow": 1048576, + "maxTokens": 1048576, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + }, + "supportsComputerUse": false + }, "minimax/minimax-m1": { "id": "minimax/minimax-m1", "name": "MiniMax M1", @@ -78547,7 +80494,7 @@ }, "minimax/minimax-m2": { "id": "minimax/minimax-m2", - "name": "MiniMax M2", + "name": "MiniMax-M2", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -78577,7 +80524,7 @@ }, "minimax/minimax-m2.1": { "id": "minimax/minimax-m2.1", - "name": "MiniMax M2.1", + "name": "MiniMax-M2.1", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -78616,7 +80563,7 @@ "text" ], "cost": { - "input": 0.15, + "input": 0.22, "output": 0.8999999999999999, "cacheRead": 0.049999999999999996, "cacheWrite": 0 @@ -78670,7 +80617,7 @@ }, "minimax/minimax-m2.7": { "id": "minimax/minimax-m2.7", - "name": "MiniMax M2.7", + "name": "MiniMax-M2.7", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -78679,9 +80626,9 @@ "text" ], "cost": { - "input": 0.25, - "output": 1, - "cacheRead": 0.049999999999999996, + "input": 0.27, + "output": 1.08, + "cacheRead": 0.054, "cacheWrite": 0 }, "contextWindow": 204800, @@ -78700,7 +80647,7 @@ }, "minimax/minimax-m3": { "id": "minimax/minimax-m3", - "name": "MiniMax M3", + "name": "MiniMax-M3", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -78975,7 +80922,7 @@ }, "mistralai/mistral-large-2512": { "id": "mistralai/mistral-large-2512", - "name": "Mistral Large 3 2512", + "name": "Mistral Large 3", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -79220,8 +81167,8 @@ "image" ], "cost": { - "input": 0.075, - "output": 0.19999999999999998, + "input": 0.09375, + "output": 0.25, "cacheRead": 0.01, "cacheWrite": 0 }, @@ -79473,9 +81420,9 @@ "image" ], "cost": { - "input": 0.6, - "output": 3.41, - "cacheRead": 0.19999999999999998, + "input": 0.5795, + "output": 2.44, + "cacheRead": 0.0976, "cacheWrite": 0 }, "contextWindow": 262144, @@ -79535,7 +81482,7 @@ "image" ], "cost": { - "input": 0.73, + "input": 0.7, "output": 3.5, "cacheRead": 0.15, "cacheWrite": 0 @@ -79554,6 +81501,36 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "moonshotai/kimi-k2.7-code:batch": { + "id": "moonshotai/kimi-k2.7-code:batch", + "name": "Kimi K2.7 Code (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.475, + "output": 2, + "cacheRead": 0.095, + "cacheWrite": 0 + }, + "contextWindow": 262144, + "maxTokens": 262144, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + }, + "supportsComputerUse": false + }, "moonshotai/kimi-k3": { "id": "moonshotai/kimi-k3", "name": "Kimi K3", @@ -79996,6 +81973,35 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "nvidia/nemotron-3-ultra-550b-a55b:batch": { + "id": "nvidia/nemotron-3-ultra-550b-a55b:batch", + "name": "Nemotron 3 Ultra (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.3, + "output": 1.7999999999999998, + "cacheRead": 0.09999999999999999, + "cacheWrite": 0 + }, + "contextWindow": 512288, + "maxTokens": null, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + }, + "supportsComputerUse": false + }, "nvidia/nemotron-3-ultra-550b-a55b:free": { "id": "nvidia/nemotron-3-ultra-550b-a55b:free", "name": "Nemotron 3 Ultra (free)", @@ -80119,7 +82125,7 @@ }, "openai/gpt-3.5-turbo": { "id": "openai/gpt-3.5-turbo", - "name": "GPT-3.5 Turbo", + "name": "GPT-3.5-turbo", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -80180,6 +82186,26 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "openai/gpt-3.5-turbo:batch": { + "id": "openai/gpt-3.5-turbo:batch", + "name": "GPT-3.5 Turbo (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": false, + "input": [ + "text" + ], + "cost": { + "input": 0.25, + "output": 0.75, + "cacheRead": 0, + "cacheWrite": 0 + }, + "contextWindow": 16385, + "maxTokens": 4096, + "supportsComputerUse": false + }, "openai/gpt-4": { "id": "openai/gpt-4", "name": "GPT-4", @@ -80286,6 +82312,27 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "openai/gpt-4-turbo:batch": { + "id": "openai/gpt-4-turbo:batch", + "name": "GPT-4 Turbo (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": false, + "input": [ + "text", + "image" + ], + "cost": { + "input": 5, + "output": 15, + "cacheRead": 0, + "cacheWrite": 0 + }, + "contextWindow": 128000, + "maxTokens": 4096, + "supportsComputerUse": false + }, "openai/gpt-4.1": { "id": "openai/gpt-4.1", "name": "GPT-4.1", @@ -80310,7 +82357,7 @@ }, "openai/gpt-4.1-mini": { "id": "openai/gpt-4.1-mini", - "name": "GPT-4.1 Mini", + "name": "GPT-4.1 mini", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -80330,9 +82377,30 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "openai/gpt-4.1-mini:batch": { + "id": "openai/gpt-4.1-mini:batch", + "name": "GPT-4.1 Mini (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": false, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.19999999999999998, + "output": 0.7999999999999999, + "cacheRead": 0.049999999999999996, + "cacheWrite": 0 + }, + "contextWindow": 1047576, + "maxTokens": 32768, + "supportsComputerUse": false + }, "openai/gpt-4.1-nano": { "id": "openai/gpt-4.1-nano", - "name": "GPT-4.1 Nano", + "name": "GPT-4.1 nano", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -80352,6 +82420,48 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "openai/gpt-4.1-nano:batch": { + "id": "openai/gpt-4.1-nano:batch", + "name": "GPT-4.1 Nano (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": false, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.049999999999999996, + "output": 0.19999999999999998, + "cacheRead": 0.012499999999999999, + "cacheWrite": 0 + }, + "contextWindow": 1047576, + "maxTokens": 32768, + "supportsComputerUse": false + }, + "openai/gpt-4.1:batch": { + "id": "openai/gpt-4.1:batch", + "name": "GPT-4.1 (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": false, + "input": [ + "text", + "image" + ], + "cost": { + "input": 1, + "output": 4, + "cacheRead": 0.25, + "cacheWrite": 0 + }, + "contextWindow": 1047576, + "maxTokens": 32768, + "supportsComputerUse": false + }, "openai/gpt-4o": { "id": "openai/gpt-4o", "name": "GPT-4o", @@ -80505,6 +82615,48 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "openai/gpt-4o-mini:batch": { + "id": "openai/gpt-4o-mini:batch", + "name": "GPT-4o-mini (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": false, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.075, + "output": 0.3, + "cacheRead": 0.0375, + "cacheWrite": 0 + }, + "contextWindow": 128000, + "maxTokens": 16384, + "supportsComputerUse": false + }, + "openai/gpt-4o:batch": { + "id": "openai/gpt-4o:batch", + "name": "GPT-4o (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": false, + "input": [ + "text", + "image" + ], + "cost": { + "input": 1.25, + "output": 5, + "cacheRead": 0.625, + "cacheWrite": 0 + }, + "contextWindow": 128000, + "maxTokens": 16384, + "supportsComputerUse": false + }, "openai/gpt-4o:extended": { "id": "openai/gpt-4o:extended", "name": "GPT-4o (extended)", @@ -80589,6 +82741,36 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "openai/gpt-5-codex:batch": { + "id": "openai/gpt-5-codex:batch", + "name": "GPT-5 Codex (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.625, + "output": 5, + "cacheRead": 0.0625, + "cacheWrite": 0 + }, + "contextWindow": 272000, + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + }, + "supportsComputerUse": false + }, "openai/gpt-5-image": { "id": "openai/gpt-5-image", "name": "GPT-5 Image", @@ -80806,6 +82988,36 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "openai/gpt-5-pro:batch": { + "id": "openai/gpt-5-pro:batch", + "name": "GPT-5 Pro (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 7.5, + "output": 60, + "cacheRead": 0, + "cacheWrite": 0 + }, + "contextWindow": 400000, + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + }, + "supportsComputerUse": false + }, "openai/gpt-5:batch": { "id": "openai/gpt-5:batch", "name": "GPT-5 (batch)", @@ -80923,7 +83135,7 @@ }, "openai/gpt-5.1-codex-max": { "id": "openai/gpt-5.1-codex-max", - "name": "GPT-5.1-Codex-Max", + "name": "GPT-5.1 Codex Max", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -80954,7 +83166,7 @@ }, "openai/gpt-5.1-codex-mini": { "id": "openai/gpt-5.1-codex-mini", - "name": "GPT-5.1-Codex-Mini", + "name": "GPT-5.1 Codex mini", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -81127,6 +83339,36 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "openai/gpt-5.2-pro:batch": { + "id": "openai/gpt-5.2-pro:batch", + "name": "GPT-5.2 Pro (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 10.5, + "output": 84, + "cacheRead": 0, + "cacheWrite": 0 + }, + "contextWindow": 400000, + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "high", + "xhigh" + ] + }, + "supportsComputerUse": false + }, "openai/gpt-5.2:batch": { "id": "openai/gpt-5.2:batch", "name": "GPT-5.2 (batch)", @@ -81244,7 +83486,7 @@ }, "openai/gpt-5.4-mini": { "id": "openai/gpt-5.4-mini", - "name": "GPT-5.4 Mini", + "name": "GPT-5.4 mini", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -81306,7 +83548,7 @@ }, "openai/gpt-5.4-nano": { "id": "openai/gpt-5.4-nano", - "name": "GPT-5.4 Nano", + "name": "GPT-5.4 nano", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -81397,6 +83639,36 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "openai/gpt-5.4-pro:batch": { + "id": "openai/gpt-5.4-pro:batch", + "name": "GPT-5.4 Pro (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 15, + "output": 90, + "cacheRead": 0, + "cacheWrite": 0 + }, + "contextWindow": 1050000, + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "high", + "xhigh" + ] + }, + "supportsComputerUse": false + }, "openai/gpt-5.4:batch": { "id": "openai/gpt-5.4:batch", "name": "GPT-5.4 (batch)", @@ -81492,6 +83764,37 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "openai/gpt-5.5-pro:batch": { + "id": "openai/gpt-5.5-pro:batch", + "name": "GPT-5.5 Pro (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 15, + "output": 90, + "cacheRead": 0, + "cacheWrite": 0 + }, + "contextWindow": 1050000, + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "high", + "xhigh" + ] + }, + "supportsComputerUse": false, + "contextPromotionTarget": "openrouter/openai/gpt-5.4" + }, "openai/gpt-5.5:batch": { "id": "openai/gpt-5.5:batch", "name": "GPT-5.5 (batch)", @@ -81537,9 +83840,9 @@ ], "cost": { "input": 0.09999999999999999, - "output": 0.6000000000000001, + "output": 0.6, "cacheRead": 0.01, - "cacheWrite": 0.12500000000000003 + "cacheWrite": 0.125 }, "contextWindow": 1050000, "maxTokens": 128000, @@ -81558,7 +83861,7 @@ }, "openai/gpt-5.6-luna-pro": { "id": "openai/gpt-5.6-luna-pro", - "name": "GPT-5.6 Luna Pro", + "name": "GPT-5.6 Luna", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -81569,9 +83872,9 @@ ], "cost": { "input": 0.09999999999999999, - "output": 0.6000000000000001, + "output": 0.6, "cacheRead": 0.01, - "cacheWrite": 0.12500000000000003 + "cacheWrite": 0.125 }, "contextWindow": 1050000, "maxTokens": 128000, @@ -81588,6 +83891,68 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "openai/gpt-5.6-luna-pro:batch": { + "id": "openai/gpt-5.6-luna-pro:batch", + "name": "GPT-5.6 Luna Pro (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.09999999999999999, + "output": 0.6, + "cacheRead": 0.01, + "cacheWrite": 0 + }, + "contextWindow": 1050000, + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "high", + "xhigh", + "max" + ] + }, + "supportsComputerUse": false + }, + "openai/gpt-5.6-luna:batch": { + "id": "openai/gpt-5.6-luna:batch", + "name": "GPT-5.6 Luna (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.09999999999999999, + "output": 0.6, + "cacheRead": 0.01, + "cacheWrite": 0 + }, + "contextWindow": 1050000, + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "high", + "xhigh", + "max" + ] + }, + "supportsComputerUse": false + }, "openai/gpt-5.6-sol": { "id": "openai/gpt-5.6-sol", "name": "GPT-5.6 Sol", @@ -81622,7 +83987,7 @@ }, "openai/gpt-5.6-sol-pro": { "id": "openai/gpt-5.6-sol-pro", - "name": "GPT-5.6 Sol Pro", + "name": "GPT-5.6 Sol", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -81652,6 +84017,68 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "openai/gpt-5.6-sol-pro:batch": { + "id": "openai/gpt-5.6-sol-pro:batch", + "name": "GPT-5.6 Sol Pro (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 2.5, + "output": 15, + "cacheRead": 0.25, + "cacheWrite": 0 + }, + "contextWindow": 1050000, + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "high", + "xhigh", + "max" + ] + }, + "supportsComputerUse": false + }, + "openai/gpt-5.6-sol:batch": { + "id": "openai/gpt-5.6-sol:batch", + "name": "GPT-5.6 Sol (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 2.5, + "output": 15, + "cacheRead": 0.25, + "cacheWrite": 0 + }, + "contextWindow": 1050000, + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "high", + "xhigh", + "max" + ] + }, + "supportsComputerUse": false + }, "openai/gpt-5.6-terra": { "id": "openai/gpt-5.6-terra", "name": "GPT-5.6 Terra", @@ -81664,7 +84091,7 @@ "image" ], "cost": { - "input": 1.0000000000000002, + "input": 1, "output": 6, "cacheRead": 0.09999999999999999, "cacheWrite": 1.25 @@ -81686,7 +84113,7 @@ }, "openai/gpt-5.6-terra-pro": { "id": "openai/gpt-5.6-terra-pro", - "name": "GPT-5.6 Terra Pro", + "name": "GPT-5.6 Terra", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -81696,7 +84123,7 @@ "image" ], "cost": { - "input": 1.0000000000000002, + "input": 1, "output": 6, "cacheRead": 0.09999999999999999, "cacheWrite": 1.25 @@ -81716,6 +84143,68 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "openai/gpt-5.6-terra-pro:batch": { + "id": "openai/gpt-5.6-terra-pro:batch", + "name": "GPT-5.6 Terra Pro (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 1, + "output": 6, + "cacheRead": 0.09999999999999999, + "cacheWrite": 0 + }, + "contextWindow": 1050000, + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "high", + "xhigh", + "max" + ] + }, + "supportsComputerUse": false + }, + "openai/gpt-5.6-terra:batch": { + "id": "openai/gpt-5.6-terra:batch", + "name": "GPT-5.6 Terra (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 1, + "output": 6, + "cacheRead": 0.09999999999999999, + "cacheWrite": 0 + }, + "contextWindow": 1050000, + "maxTokens": 128000, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "medium", + "high", + "xhigh", + "max" + ] + }, + "supportsComputerUse": false + }, "openai/gpt-audio": { "id": "openai/gpt-audio", "name": "GPT Audio", @@ -81995,6 +84484,37 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "openai/o1:batch": { + "id": "openai/o1:batch", + "name": "o1 (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 7.5, + "output": 30, + "cacheRead": 3.75, + "cacheWrite": 0 + }, + "contextWindow": 200000, + "maxTokens": 100000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ], + "requiresEffort": true + }, + "supportsComputerUse": false + }, "openai/o3": { "id": "openai/o3", "name": "o3", @@ -82096,7 +84616,7 @@ "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", - "reasoning": false, + "reasoning": true, "input": [ "text" ], @@ -82109,7 +84629,77 @@ "contextWindow": 200000, "maxTokens": 100000, "supportsComputerUse": false, - "supportsComputerUseConfig": false + "supportsComputerUseConfig": false, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ], + "requiresEffort": true + } + }, + "openai/o3-mini-high:batch": { + "id": "openai/o3-mini-high:batch", + "name": "o3 Mini High (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.55, + "output": 2.2, + "cacheRead": 0.275, + "cacheWrite": 0 + }, + "contextWindow": 200000, + "maxTokens": 100000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ], + "requiresEffort": true + }, + "supportsComputerUse": false + }, + "openai/o3-mini:batch": { + "id": "openai/o3-mini:batch", + "name": "o3 Mini (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.55, + "output": 2.2, + "cacheRead": 0.275, + "cacheWrite": 0 + }, + "contextWindow": 200000, + "maxTokens": 100000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ], + "requiresEffort": true + }, + "supportsComputerUse": false }, "openai/o3-pro": { "id": "openai/o3-pro", @@ -82143,6 +84733,68 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "openai/o3-pro:batch": { + "id": "openai/o3-pro:batch", + "name": "o3 Pro (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 10, + "output": 40, + "cacheRead": 0, + "cacheWrite": 0 + }, + "contextWindow": 200000, + "maxTokens": 100000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ], + "requiresEffort": true + }, + "supportsComputerUse": false + }, + "openai/o3:batch": { + "id": "openai/o3:batch", + "name": "o3 (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 1, + "output": 4, + "cacheRead": 0.25, + "cacheWrite": 0 + }, + "contextWindow": 200000, + "maxTokens": 100000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ], + "requiresEffort": true + }, + "supportsComputerUse": false + }, "openai/o4-mini": { "id": "openai/o4-mini", "name": "o4-mini", @@ -82239,6 +84891,68 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "openai/o4-mini-high:batch": { + "id": "openai/o4-mini-high:batch", + "name": "o4 Mini High (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.55, + "output": 2.2, + "cacheRead": 0.1375, + "cacheWrite": 0 + }, + "contextWindow": 200000, + "maxTokens": 100000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ], + "requiresEffort": true + }, + "supportsComputerUse": false + }, + "openai/o4-mini:batch": { + "id": "openai/o4-mini:batch", + "name": "o4 Mini (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.55, + "output": 2.2, + "cacheRead": 0.1375, + "cacheWrite": 0 + }, + "contextWindow": 200000, + "maxTokens": 100000, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ], + "requiresEffort": true + }, + "supportsComputerUse": false + }, "openrouter/aurora-alpha": { "id": "openrouter/aurora-alpha", "name": "Aurora Alpha", @@ -82354,7 +85068,7 @@ }, "openrouter/free": { "id": "openrouter/free", - "name": "Free Models Router", + "name": "OpenRouter Free Models Router", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -82812,7 +85526,7 @@ }, "qwen/qwen-plus": { "id": "qwen/qwen-plus", - "name": "Qwen-Plus", + "name": "Qwen Plus", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -82957,7 +85671,7 @@ }, "qwen/qwen3-235b-a22b": { "id": "qwen/qwen3-235b-a22b", - "name": "Qwen3 235B A22B", + "name": "Qwen3 235B-A22B", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -83121,7 +85835,7 @@ }, "qwen/qwen3-32b": { "id": "qwen/qwen3-32b", - "name": "Qwen3-32B", + "name": "Qwen3 32B", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -83262,7 +85976,7 @@ }, "qwen/qwen3-coder-30b-a3b-instruct": { "id": "qwen/qwen3-coder-30b-a3b-instruct", - "name": "Qwen3 Coder 30B A3B Instruct", + "name": "Qwen3-Coder 30B-A3B Instruct", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -83272,12 +85986,12 @@ ], "cost": { "input": 0.07, - "output": 0.28, + "output": 0.27, "cacheRead": 0, "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 262144, + "maxTokens": 32768, "supportsComputerUse": false, "supportsComputerUseConfig": false }, @@ -83449,13 +86163,13 @@ "text" ], "cost": { - "input": 0.09999999999999999, + "input": 0.09, "output": 1.1, "cacheRead": 0.07, "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 262144, + "maxTokens": 16384, "supportsComputerUse": false, "supportsComputerUseConfig": false }, @@ -83482,7 +86196,7 @@ }, "qwen/qwen3-next-80b-a3b-thinking": { "id": "qwen/qwen3-next-80b-a3b-thinking", - "name": "Qwen3 Next 80B A3B Thinking", + "name": "Qwen3-Next 80B-A3B (Thinking)", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -83497,7 +86211,7 @@ "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 32768, + "maxTokens": 262144, "thinking": { "mode": "effort", "efforts": [ @@ -83577,13 +86291,13 @@ "image" ], "cost": { - "input": 0.13, - "output": 0.52, + "input": 0.15, + "output": 0.6, "cacheRead": 0, "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 32768, + "maxTokens": 16384, "supportsComputerUse": false, "supportsComputerUseConfig": false }, @@ -83707,13 +86421,13 @@ "image" ], "cost": { - "input": 0.26, - "output": 2.08, + "input": 0.29, + "output": 2.4, "cacheRead": 0, "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 65536, + "maxTokens": 81920, "thinking": { "mode": "effort", "efforts": [ @@ -83728,7 +86442,7 @@ }, "qwen/qwen3.5-27b": { "id": "qwen/qwen3.5-27b", - "name": "Qwen3.5-27B", + "name": "Qwen3.5 27B", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -83759,7 +86473,7 @@ }, "qwen/qwen3.5-35b-a3b": { "id": "qwen/qwen3.5-35b-a3b", - "name": "Qwen3.5-35B-A3B", + "name": "Qwen3.5 35B-A3B", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -83790,7 +86504,7 @@ }, "qwen/qwen3.5-397b-a17b": { "id": "qwen/qwen3.5-397b-a17b", - "name": "Qwen3.5 397B A17B", + "name": "Qwen3.5 397B-A17B", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -83821,7 +86535,7 @@ }, "qwen/qwen3.5-9b": { "id": "qwen/qwen3.5-9b", - "name": "Qwen3.5-9B", + "name": "Qwen3.5 9B", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -83955,13 +86669,13 @@ "image" ], "cost": { - "input": 0.3, - "output": 2, - "cacheRead": 0.15, + "input": 0.6, + "output": 3.5999999999999996, + "cacheRead": 0.12, "cacheWrite": 0 }, "contextWindow": 262144, - "maxTokens": 65536, + "maxTokens": 262144, "thinking": { "mode": "effort", "efforts": [ @@ -83976,7 +86690,7 @@ }, "qwen/qwen3.6-35b-a3b": { "id": "qwen/qwen3.6-35b-a3b", - "name": "Qwen3.6 35B A3B", + "name": "Qwen3.6 35B-A3B", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -84250,6 +86964,36 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "qwen/qwen3.8-max": { + "id": "qwen/qwen3.8-max", + "name": "Qwen3.8 Max", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 2, + "output": 6, + "cacheRead": 0.25, + "cacheWrite": 2.5 + }, + "contextWindow": 1000000, + "maxTokens": 131072, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + }, + "supportsComputerUse": false + }, "qwen/qwq-32b": { "id": "qwen/qwq-32b", "name": "QwQ 32B", @@ -84520,7 +87264,7 @@ }, "tencent/hy3": { "id": "tencent/hy3", - "name": "Hy3", + "name": "Tencent Hy3", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -84550,7 +87294,7 @@ }, "tencent/hy3-preview": { "id": "tencent/hy3-preview", - "name": "Hy3 Preview", + "name": "Hy3 preview", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -84676,7 +87420,7 @@ "cacheWrite": 0 }, "contextWindow": 1024000, - "maxTokens": 32768, + "maxTokens": 1024000, "supportsComputerUse": false, "supportsComputerUseConfig": false }, @@ -84692,13 +87436,13 @@ "image" ], "cost": { - "input": 1, + "input": 0.95, "output": 4.05, - "cacheRead": 0.16999999999999998, + "cacheRead": 0.16, "cacheWrite": 0 }, "contextWindow": 1048576, - "maxTokens": 65536, + "maxTokens": 262144, "thinking": { "mode": "effort", "efforts": [ @@ -84723,13 +87467,13 @@ "image" ], "cost": { - "input": 0.5, + "input": 0.44999999999999996, "output": 1.2, "cacheRead": 0.09999999999999999, "cacheWrite": 0 }, "contextWindow": 524288, - "maxTokens": 32768, + "maxTokens": 262144, "thinking": { "mode": "effort", "efforts": [ @@ -84742,6 +87486,36 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "thinkingmachines/inkling:batch": { + "id": "thinkingmachines/inkling:batch", + "name": "Inkling (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.5, + "output": 2.025, + "cacheRead": 0.08499999999999999, + "cacheWrite": 0 + }, + "contextWindow": 524288, + "maxTokens": null, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + }, + "supportsComputerUse": false + }, "tngtech/deepseek-r1t2-chimera": { "id": "tngtech/deepseek-r1t2-chimera", "name": "DeepSeek R1T2 Chimera", @@ -84815,8 +87589,8 @@ "cacheRead": 0.015, "cacheWrite": 0 }, - "contextWindow": 128000, - "maxTokens": 32768, + "contextWindow": 131072, + "maxTokens": 131072, "thinking": { "mode": "effort", "efforts": [ @@ -85300,7 +88074,7 @@ }, "xiaomi/mimo-v2-pro": { "id": "xiaomi/mimo-v2-pro", - "name": "MiMo-V2-Pro", + "name": "MiMo V2 Pro", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -85359,7 +88133,7 @@ }, "xiaomi/mimo-v2.5-pro": { "id": "xiaomi/mimo-v2.5-pro", - "name": "MiMo V2.5 Pro", + "name": "MiMo-V2.5-Pro", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -85409,7 +88183,7 @@ }, "z-ai/glm-4.5": { "id": "z-ai/glm-4.5", - "name": "GLM 4.5", + "name": "GLM-4.5", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -85439,7 +88213,7 @@ }, "z-ai/glm-4.5-air": { "id": "z-ai/glm-4.5-air", - "name": "GLM 4.5 Air", + "name": "GLM-4.5-Air", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -85499,7 +88273,7 @@ }, "z-ai/glm-4.5v": { "id": "z-ai/glm-4.5v", - "name": "GLM 4.5V", + "name": "GLM-4.5V", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -85530,7 +88304,7 @@ }, "z-ai/glm-4.6": { "id": "z-ai/glm-4.6", - "name": "GLM 4.6", + "name": "GLM-4.6", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -85621,7 +88395,7 @@ }, "z-ai/glm-4.7": { "id": "z-ai/glm-4.7", - "name": "GLM 4.7", + "name": "GLM-4.7", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -85651,7 +88425,7 @@ }, "z-ai/glm-4.7-flash": { "id": "z-ai/glm-4.7-flash", - "name": "GLM 4.7 Flash", + "name": "GLM-4.7-Flash", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -85681,7 +88455,7 @@ }, "z-ai/glm-5": { "id": "z-ai/glm-5", - "name": "GLM 5", + "name": "GLM-5", "api": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", "provider": "openrouter", @@ -85732,7 +88506,7 @@ }, "z-ai/glm-5.1": { "id": "z-ai/glm-5.1", - "name": "GLM 5.1", + "name": "GLM-5.1", "api": "openrouter", "provider": "openrouter", "baseUrl": "https://openrouter.ai/api/v1", @@ -85741,13 +88515,13 @@ "text" ], "cost": { - "input": 0.966, - "output": 3.036, - "cacheRead": 0.1794, + "input": 0.952, + "output": 2.992, + "cacheRead": 0.17679999999999998, "cacheWrite": 0 }, "contextWindow": 204800, - "maxTokens": 128000, + "maxTokens": 131072, "thinking": { "mode": "effort", "efforts": [ @@ -85771,9 +88545,9 @@ "text" ], "cost": { - "input": 0.42, - "output": 1.32, - "cacheRead": 0.078, + "input": 0.2212, + "output": 0.6952, + "cacheRead": 0.04108, "cacheWrite": 0 }, "contextWindow": 1048576, @@ -85791,6 +88565,36 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "z-ai/glm-5.2:batch": { + "id": "z-ai/glm-5.2:batch", + "name": "GLM 5.2 (batch)", + "api": "openrouter", + "provider": "openrouter", + "baseUrl": "https://openrouter.ai/api/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.7, + "output": 2.2, + "cacheRead": 0.13, + "cacheWrite": 0 + }, + "contextWindow": 512000, + "maxTokens": 131072, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + }, + "supportsComputerUse": false + }, "z-ai/glm-5v-turbo": { "id": "z-ai/glm-5v-turbo", "name": "GLM 5V Turbo", @@ -85976,9 +88780,69 @@ } }, "synthetic": { + "hf:MiniMaxAI/MiniMax-M3": { + "id": "hf:MiniMaxAI/MiniMax-M3", + "name": "MiniMax-M3", + "api": "openai-completions", + "provider": "synthetic", + "baseUrl": "https://api.synthetic.new/openai/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.6, + "output": 1.2, + "cacheRead": 0.6, + "cacheWrite": 0 + }, + "contextWindow": 524288, + "maxTokens": 65536, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } + }, + "hf:moonshotai/Kimi-K2.7-Code": { + "id": "hf:moonshotai/Kimi-K2.7-Code", + "name": "Kimi K2.7 Code", + "api": "openai-completions", + "provider": "synthetic", + "baseUrl": "https://api.synthetic.new/openai/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.95, + "output": 4, + "cacheRead": 0.95, + "cacheWrite": 0 + }, + "contextWindow": 262144, + "maxTokens": 65536, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } + }, "hf:moonshotai/Kimi-K3": { "id": "hf:moonshotai/Kimi-K3", - "name": "moonshotai/Kimi-K3", + "name": "Kimi K3", "api": "openai-completions", "provider": "synthetic", "baseUrl": "https://api.synthetic.new/openai/v1", @@ -86007,12 +88871,11 @@ "max": "max" }, "requiresEffort": true - }, - "supportsComputerUse": false + } }, "hf:nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4": { "id": "hf:nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4", - "name": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4", + "name": "Nemotron 3 Super 120B A12B", "api": "openai-completions", "provider": "synthetic", "baseUrl": "https://api.synthetic.new/openai/v1", @@ -86023,7 +88886,7 @@ "cost": { "input": 0.3, "output": 1, - "cacheRead": 0.06, + "cacheRead": 0.3, "cacheWrite": 0 }, "contextWindow": 262144, @@ -86037,12 +88900,11 @@ "high", "xhigh" ] - }, - "supportsComputerUse": false + } }, "hf:openai/gpt-oss-120b": { "id": "hf:openai/gpt-oss-120b", - "name": "openai/gpt-oss-120b", + "name": "GPT OSS 120B", "api": "openai-completions", "provider": "synthetic", "baseUrl": "https://api.synthetic.new/openai/v1", @@ -86053,11 +88915,11 @@ "cost": { "input": 0.1, "output": 0.1, - "cacheRead": 0.02, + "cacheRead": 0.1, "cacheWrite": 0 }, "contextWindow": 131072, - "maxTokens": 65536, + "maxTokens": 32768, "thinking": { "mode": "effort", "efforts": [ @@ -86065,12 +88927,11 @@ "medium", "high" ] - }, - "supportsComputerUse": false + } }, "hf:Qwen/Qwen3.6-27B": { "id": "hf:Qwen/Qwen3.6-27B", - "name": "Qwen/Qwen3.6-27B", + "name": "Qwen3.6 27B", "api": "openai-completions", "provider": "synthetic", "baseUrl": "https://api.synthetic.new/openai/v1", @@ -86081,8 +88942,8 @@ ], "cost": { "input": 0.45, - "output": 2.2, - "cacheRead": 0.09, + "output": 3.6, + "cacheRead": 0.45, "cacheWrite": 0 }, "contextWindow": 262144, @@ -86095,12 +88956,11 @@ "medium", "high" ] - }, - "supportsComputerUse": false + } }, "hf:zai-org/GLM-4.7-Flash": { "id": "hf:zai-org/GLM-4.7-Flash", - "name": "zai-org/GLM-4.7-Flash", + "name": "GLM-4.7-Flash", "api": "openai-completions", "provider": "synthetic", "baseUrl": "https://api.synthetic.new/openai/v1", @@ -86111,7 +88971,7 @@ "cost": { "input": 0.1, "output": 0.5, - "cacheRead": 0.02, + "cacheRead": 0.1, "cacheWrite": 0 }, "contextWindow": 196608, @@ -86125,12 +88985,11 @@ "high", "xhigh" ] - }, - "supportsComputerUse": false + } }, "hf:zai-org/GLM-5.2": { "id": "hf:zai-org/GLM-5.2", - "name": "zai-org/GLM-5.2", + "name": "GLM-5.2", "api": "openai-completions", "provider": "synthetic", "baseUrl": "https://api.synthetic.new/openai/v1", @@ -86139,9 +88998,9 @@ "text" ], "cost": { - "input": 1, - "output": 3, - "cacheRead": 0.16, + "input": 1.4, + "output": 4.4, + "cacheRead": 1.4, "cacheWrite": 0 }, "contextWindow": 524288, @@ -86155,8 +89014,7 @@ "high", "xhigh" ] - }, - "supportsComputerUse": false + } }, "syn:large:text": { "id": "syn:large:text", @@ -86376,6 +89234,33 @@ "contextWindow": 131072, "maxTokens": 8192 }, + "deepseek-ai/DeepSeek-V4-Flash-0731": { + "id": "deepseek-ai/DeepSeek-V4-Flash-0731", + "name": "DeepSeek V4 Flash 0731", + "api": "openai-completions", + "provider": "together", + "baseUrl": "https://api.together.xyz/v1", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.14, + "output": 0.28, + "cacheRead": 0.03, + "cacheWrite": 0 + }, + "contextWindow": 1000000, + "maxTokens": 384000, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "high", + "max" + ] + } + }, "deepseek-ai/DeepSeek-V4-Pro": { "id": "deepseek-ai/DeepSeek-V4-Pro", "name": "DeepSeek V4 Pro", @@ -87187,7 +90072,41 @@ }, "umans-deepseek-v4-flash-0731": { "id": "umans-deepseek-v4-flash-0731", - "name": "Umans DeepSeek V4 Flash (experimental)", + "name": "Umans DeepSeek V4 Flash", + "api": "anthropic-messages", + "provider": "umans", + "baseUrl": "https://api.code.umans.ai", + "reasoning": true, + "thinking": { + "mode": "budget", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + }, + "input": [ + "text" + ], + "cost": { + "input": 0.14, + "output": 0.28, + "cacheRead": 0.028, + "cacheWrite": 0 + }, + "contextWindow": 1048576, + "maxTokens": 393215, + "supportsComputerUse": false, + "supportsComputerUseConfig": false, + "compat": { + "escapeBuiltinToolNames": true + } + }, + "umans-deepseek-v4-flash-0731-lab": { + "id": "umans-deepseek-v4-flash-0731-lab", + "name": "Umans DeepSeek V4 Flash (lab)", "api": "anthropic-messages", "provider": "umans", "baseUrl": "https://api.code.umans.ai", @@ -87214,7 +90133,6 @@ "contextWindow": 1048576, "maxTokens": 393215, "supportsComputerUse": false, - "supportsComputerUseConfig": false, "compat": { "escapeBuiltinToolNames": true } @@ -87343,9 +90261,9 @@ "image" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 3, + "output": 15, + "cacheRead": 0.3, "cacheWrite": 0 }, "contextWindow": 1048576, @@ -88023,7 +90941,7 @@ }, "deepseek-v4-flash": { "id": "deepseek-v4-flash", - "name": "DeepSeek V4 Flash", + "name": "DeepSeek V4 Flash 0423", "api": "openai-completions", "provider": "venice", "baseUrl": "https://api.venice.ai/api/v1", @@ -88042,6 +90960,7 @@ "thinking": { "mode": "effort", "efforts": [ + "low", "high", "max" ] @@ -88068,6 +90987,7 @@ "thinking": { "mode": "effort", "efforts": [ + "low", "high", "max" ] @@ -89359,6 +92279,39 @@ "requiresEffort": true } }, + "kimi-k3-fast-api": { + "id": "kimi-k3-fast-api", + "name": "Kimi K3 Fast", + "api": "openai-completions", + "provider": "venice", + "baseUrl": "https://api.venice.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 4.5, + "output": 22.5, + "cacheRead": 0.45, + "cacheWrite": 0 + }, + "contextWindow": 1000000, + "maxTokens": 131072, + "thinking": { + "mode": "effort", + "efforts": [ + "low", + "high", + "max" + ], + "defaultLevel": "max", + "effortMap": { + "max": "max" + }, + "requiresEffort": true + } + }, "llama-3.2-3b": { "id": "llama-3.2-3b", "name": "Llama 3.2 3B", @@ -90334,6 +93287,35 @@ ] } }, + "qwen-3-8-max": { + "id": "qwen-3-8-max", + "name": "Qwen 3.8 Max", + "api": "openai-completions", + "provider": "venice", + "baseUrl": "https://api.venice.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 2.5, + "output": 7.5, + "cacheRead": 0.3125, + "cacheWrite": 3.125 + }, + "contextWindow": 1000000, + "maxTokens": 131072, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + } + }, "qwen3-235b-a22b-instruct-2507": { "id": "qwen3-235b-a22b-instruct-2507", "name": "Qwen 3 235B A22B Instruct 2507", @@ -91726,6 +94708,37 @@ }, "supportsComputerUse": false }, + "alibaba/qwen3.8-max": { + "id": "alibaba/qwen3.8-max", + "name": "Qwen 3.8 Max", + "api": "anthropic-messages", + "provider": "vercel-ai-gateway", + "baseUrl": "https://ai-gateway.vercel.sh", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 2, + "output": 6, + "cacheRead": 0.25, + "cacheWrite": 2.5 + }, + "contextWindow": 1000000, + "maxTokens": 128000, + "thinking": { + "mode": "budget", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + }, + "supportsComputerUse": false + }, "amazon/nova-2-lite": { "id": "amazon/nova-2-lite", "name": "Nova 2 Lite", @@ -92061,7 +95074,8 @@ "xhigh" ] }, - "supportsComputerUse": false + "supportsComputerUse": false, + "supportsComputerUseConfig": false }, "anthropic/claude-opus-4.5": { "id": "anthropic/claude-opus-4.5", @@ -92158,7 +95172,7 @@ }, "anthropic/claude-opus-4.7-fast": { "id": "anthropic/claude-opus-4.7-fast", - "name": "Claude Opus 4.7 (Fast)", + "name": "Claude Opus 4.7", "api": "anthropic-messages", "provider": "vercel-ai-gateway", "baseUrl": "https://ai-gateway.vercel.sh", @@ -92223,7 +95237,7 @@ }, "anthropic/claude-opus-4.8-fast": { "id": "anthropic/claude-opus-4.8-fast", - "name": "Claude Opus 4.8 (Fast)", + "name": "Claude Opus 4.8", "api": "anthropic-messages", "provider": "vercel-ai-gateway", "baseUrl": "https://ai-gateway.vercel.sh", @@ -92287,7 +95301,7 @@ }, "anthropic/claude-opus-5-fast": { "id": "anthropic/claude-opus-5-fast", - "name": "Claude Opus 5 (Fast)", + "name": "Claude Opus 5", "api": "anthropic-messages", "provider": "vercel-ai-gateway", "baseUrl": "https://ai-gateway.vercel.sh", @@ -92315,7 +95329,8 @@ ], "supportsDisplay": true }, - "supportsComputerUse": false + "supportsComputerUse": false, + "supportsComputerUseConfig": false }, "anthropic/claude-sonnet-4": { "id": "anthropic/claude-sonnet-4", @@ -92606,7 +95621,7 @@ }, "deepseek/deepseek-r1": { "id": "deepseek/deepseek-r1", - "name": "R1", + "name": "DeepSeek-R1", "api": "anthropic-messages", "provider": "vercel-ai-gateway", "baseUrl": "https://ai-gateway.vercel.sh", @@ -92787,9 +95802,9 @@ "text" ], "cost": { - "input": 0.14, - "output": 0.28, - "cacheRead": 0.0028, + "input": 0.19999999999999998, + "output": 0.39999999999999997, + "cacheRead": 0.04, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -92817,9 +95832,9 @@ "text" ], "cost": { - "input": 0.13, - "output": 0.26, - "cacheRead": 0.028, + "input": 0.19999999999999998, + "output": 0.39999999999999997, + "cacheRead": 0.04, "cacheWrite": 0 }, "contextWindow": 1000000, @@ -92847,13 +95862,13 @@ "text" ], "cost": { - "input": 0.435, - "output": 0.87, - "cacheRead": 0.0036, + "input": 1.74, + "output": 3.48, + "cacheRead": 0.14, "cacheWrite": 0 }, - "contextWindow": 1000000, - "maxTokens": 384000, + "contextWindow": 1048600, + "maxTokens": 1048600, "thinking": { "mode": "budget", "efforts": [ @@ -92942,7 +95957,7 @@ }, "google/gemini-2.5-flash-lite": { "id": "google/gemini-2.5-flash-lite", - "name": "Gemini 2.5 Flash Lite", + "name": "Gemini 2.5 Flash-Lite", "api": "anthropic-messages", "baseUrl": "https://ai-gateway.vercel.sh", "provider": "vercel-ai-gateway", @@ -93301,7 +96316,7 @@ }, "google/gemma-4-26b-a4b-it": { "id": "google/gemma-4-26b-a4b-it", - "name": "Gemma 4 26B A4B", + "name": "Gemma 4 26B A4B IT", "api": "anthropic-messages", "provider": "vercel-ai-gateway", "baseUrl": "https://ai-gateway.vercel.sh", @@ -93332,7 +96347,7 @@ }, "google/gemma-4-31b-it": { "id": "google/gemma-4-31b-it", - "name": "Gemma 4 31B", + "name": "Gemma 4 31B IT", "api": "anthropic-messages", "provider": "vercel-ai-gateway", "baseUrl": "https://ai-gateway.vercel.sh", @@ -93347,7 +96362,7 @@ "cacheRead": 0, "cacheWrite": 0 }, - "contextWindow": 262144, + "contextWindow": 256000, "maxTokens": 131072, "thinking": { "mode": "budget", @@ -93411,6 +96426,36 @@ "maxTokens": 16384, "supportsComputerUse": false }, + "inclusionai/ling-3.0-flash": { + "id": "inclusionai/ling-3.0-flash", + "name": "Ling-3.0-flash", + "api": "anthropic-messages", + "provider": "vercel-ai-gateway", + "baseUrl": "https://ai-gateway.vercel.sh", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0.06, + "output": 0.18, + "cacheRead": 0.012, + "cacheWrite": 0 + }, + "contextWindow": 256000, + "maxTokens": 32000, + "thinking": { + "mode": "budget", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + }, + "supportsComputerUse": false + }, "inclusionai/ling-3.0-flash-free": { "id": "inclusionai/ling-3.0-flash-free", "name": "Ling 3.0 Flash", @@ -93439,6 +96484,37 @@ "xhigh" ] }, + "supportsComputerUse": false, + "supportsComputerUseConfig": false + }, + "inclusionai/ling-3.0-tiny-free": { + "id": "inclusionai/ling-3.0-tiny-free", + "name": "Ling 3.0 Tiny (Free)", + "api": "anthropic-messages", + "provider": "vercel-ai-gateway", + "baseUrl": "https://ai-gateway.vercel.sh", + "reasoning": true, + "input": [ + "text" + ], + "cost": { + "input": 0, + "output": 0, + "cacheRead": 0, + "cacheWrite": 0 + }, + "contextWindow": 256000, + "maxTokens": 32000, + "thinking": { + "mode": "budget", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + }, "supportsComputerUse": false }, "interfaze/interfaze-beta": { @@ -93474,14 +96550,13 @@ }, "kwaipilot/kat-coder-air-v2.5": { "id": "kwaipilot/kat-coder-air-v2.5", - "name": "Kat Coder Air V2.5", + "name": "KAT-Coder-Air V2.5", "api": "anthropic-messages", "provider": "vercel-ai-gateway", "baseUrl": "https://ai-gateway.vercel.sh", - "reasoning": true, + "reasoning": false, "input": [ - "text", - "image" + "text" ], "cost": { "input": 0.15, @@ -93491,16 +96566,6 @@ }, "contextWindow": 256000, "maxTokens": 80000, - "thinking": { - "mode": "budget", - "efforts": [ - "minimal", - "low", - "medium", - "high", - "xhigh" - ] - }, "supportsComputerUse": false }, "kwaipilot/kat-coder-pro-v1": { @@ -93545,14 +96610,13 @@ }, "kwaipilot/kat-coder-pro-v2.5": { "id": "kwaipilot/kat-coder-pro-v2.5", - "name": "Kat Coder Pro V2.5", + "name": "KAT-Coder-Pro V2.5", "api": "anthropic-messages", "provider": "vercel-ai-gateway", "baseUrl": "https://ai-gateway.vercel.sh", - "reasoning": true, + "reasoning": false, "input": [ - "text", - "image" + "text" ], "cost": { "input": 0.74, @@ -93562,16 +96626,6 @@ }, "contextWindow": 256000, "maxTokens": 80000, - "thinking": { - "mode": "budget", - "efforts": [ - "minimal", - "low", - "medium", - "high", - "xhigh" - ] - }, "supportsComputerUse": false }, "meituan/longcat-flash-chat": { @@ -93836,9 +96890,71 @@ }, "supportsComputerUse": false }, + "meta/muse-spark-1.2": { + "id": "meta/muse-spark-1.2", + "name": "Muse Spark 1.2", + "api": "anthropic-messages", + "provider": "vercel-ai-gateway", + "baseUrl": "https://ai-gateway.vercel.sh", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 1.25, + "output": 4.25, + "cacheRead": 0.15, + "cacheWrite": 0 + }, + "contextWindow": 1048576, + "maxTokens": 1048576, + "thinking": { + "mode": "budget", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + }, + "supportsComputerUse": false + }, + "meta/muse-spark-1.2-contributor": { + "id": "meta/muse-spark-1.2-contributor", + "name": "Muse Spark 1.2 Contributor (Data Used for Training)", + "api": "anthropic-messages", + "provider": "vercel-ai-gateway", + "baseUrl": "https://ai-gateway.vercel.sh", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.09999999999999999, + "output": 0.19999999999999998, + "cacheRead": 0.002, + "cacheWrite": 0 + }, + "contextWindow": 1048576, + "maxTokens": 1048576, + "thinking": { + "mode": "budget", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + }, + "supportsComputerUse": false + }, "minimax/minimax-m2": { "id": "minimax/minimax-m2", - "name": "MiniMax M2", + "name": "MiniMax-M2", "api": "anthropic-messages", "baseUrl": "https://ai-gateway.vercel.sh", "provider": "vercel-ai-gateway", @@ -93872,7 +96988,7 @@ }, "minimax/minimax-m2.1": { "id": "minimax/minimax-m2.1", - "name": "MiniMax M2.1", + "name": "MiniMax-M2.1", "api": "anthropic-messages", "baseUrl": "https://ai-gateway.vercel.sh", "provider": "vercel-ai-gateway", @@ -94009,7 +97125,7 @@ }, "minimax/minimax-m2.7": { "id": "minimax/minimax-m2.7", - "name": "MiniMax M2.7", + "name": "MiniMax-M2.7", "api": "anthropic-messages", "provider": "vercel-ai-gateway", "baseUrl": "https://ai-gateway.vercel.sh", @@ -94077,7 +97193,7 @@ }, "minimax/minimax-m3": { "id": "minimax/minimax-m3", - "name": "MiniMax M3", + "name": "MiniMax-M3", "api": "anthropic-messages", "provider": "vercel-ai-gateway", "baseUrl": "https://ai-gateway.vercel.sh", @@ -94683,7 +97799,7 @@ }, "moonshotai/kimi-k2.7-code-highspeed": { "id": "moonshotai/kimi-k2.7-code-highspeed", - "name": "Kimi K2.7 Code High Speed", + "name": "Kimi K2.7 Code High-Speed", "api": "anthropic-messages", "provider": "vercel-ai-gateway", "baseUrl": "https://ai-gateway.vercel.sh", @@ -94897,7 +98013,7 @@ }, "nvidia/nemotron-nano-9b-v2": { "id": "nvidia/nemotron-nano-9b-v2", - "name": "Nemotron Nano 9B V2", + "name": "Nvidia Nemotron Nano 9B V2", "api": "anthropic-messages", "baseUrl": "https://ai-gateway.vercel.sh", "provider": "vercel-ai-gateway", @@ -94959,7 +98075,7 @@ }, "openai/gpt-3.5-turbo": { "id": "openai/gpt-3.5-turbo", - "name": "GPT-3.5 Turbo", + "name": "GPT-3.5-turbo", "api": "anthropic-messages", "provider": "vercel-ai-gateway", "baseUrl": "https://ai-gateway.vercel.sh", @@ -95021,7 +98137,7 @@ }, "openai/gpt-4.1-mini": { "id": "openai/gpt-4.1-mini", - "name": "GPT-4.1 Mini", + "name": "GPT-4.1 mini", "api": "anthropic-messages", "baseUrl": "https://ai-gateway.vercel.sh", "provider": "vercel-ai-gateway", @@ -95042,7 +98158,7 @@ }, "openai/gpt-4.1-nano": { "id": "openai/gpt-4.1-nano", - "name": "GPT-4.1 Nano", + "name": "GPT-4.1 nano", "api": "anthropic-messages", "baseUrl": "https://ai-gateway.vercel.sh", "provider": "vercel-ai-gateway", @@ -95316,7 +98432,7 @@ }, "openai/gpt-5.1-codex-max": { "id": "openai/gpt-5.1-codex-max", - "name": "GPT-5.1-Codex-Max", + "name": "GPT-5.1 Codex Max", "api": "anthropic-messages", "baseUrl": "https://ai-gateway.vercel.sh", "provider": "vercel-ai-gateway", @@ -95346,7 +98462,7 @@ }, "openai/gpt-5.1-codex-mini": { "id": "openai/gpt-5.1-codex-mini", - "name": "GPT-5.1-Codex-Mini", + "name": "GPT-5.1 Codex mini", "api": "anthropic-messages", "baseUrl": "https://ai-gateway.vercel.sh", "provider": "vercel-ai-gateway", @@ -95628,7 +98744,7 @@ }, "openai/gpt-5.4-mini": { "id": "openai/gpt-5.4-mini", - "name": "GPT-5.4 Mini", + "name": "GPT-5.4 mini", "api": "anthropic-messages", "provider": "vercel-ai-gateway", "baseUrl": "https://ai-gateway.vercel.sh", @@ -95658,7 +98774,7 @@ }, "openai/gpt-5.4-nano": { "id": "openai/gpt-5.4-nano", - "name": "GPT-5.4 Nano", + "name": "GPT-5.4 nano", "api": "anthropic-messages", "provider": "vercel-ai-gateway", "baseUrl": "https://ai-gateway.vercel.sh", @@ -96027,7 +99143,7 @@ }, "openai/o3-deep-research": { "id": "openai/o3-deep-research", - "name": "o3 Deep Research", + "name": "o3-deep-research", "api": "anthropic-messages", "baseUrl": "https://ai-gateway.vercel.sh", "provider": "vercel-ai-gateway", @@ -96381,7 +99497,7 @@ }, "tencent/hy3": { "id": "tencent/hy3", - "name": "Hy3", + "name": "Tencent Hy3", "api": "anthropic-messages", "provider": "vercel-ai-gateway", "baseUrl": "https://ai-gateway.vercel.sh", @@ -97086,7 +100202,7 @@ }, "xiaomi/mimo-v2-pro": { "id": "xiaomi/mimo-v2-pro", - "name": "MiMo-V2-Pro", + "name": "MiMo V2 Pro", "api": "anthropic-messages", "provider": "vercel-ai-gateway", "baseUrl": "https://ai-gateway.vercel.sh", @@ -97148,7 +100264,7 @@ }, "xiaomi/mimo-v2.5-pro": { "id": "xiaomi/mimo-v2.5-pro", - "name": "MiMo V2.5 Pro", + "name": "MiMo-V2.5-Pro", "api": "anthropic-messages", "provider": "vercel-ai-gateway", "baseUrl": "https://ai-gateway.vercel.sh", @@ -97526,8 +100642,8 @@ "cacheRead": 0.26, "cacheWrite": 0 }, - "contextWindow": 202000, - "maxTokens": 202000, + "contextWindow": 202800, + "maxTokens": 64000, "thinking": { "mode": "budget", "efforts": [ @@ -97654,6 +100770,7 @@ "thinking": { "mode": "effort", "efforts": [ + "low", "high", "max" ] @@ -98872,7 +101989,6 @@ }, "contextWindow": 2000000, "maxTokens": 2000000, - "supportsComputerUse": false, "compat": { "reasoningEffortMap": { "minimal": "low" @@ -98903,7 +102019,6 @@ }, "contextWindow": 2000000, "maxTokens": 2000000, - "supportsComputerUse": false, "compat": { "reasoningEffortMap": { "minimal": "low" @@ -98933,6 +102048,16 @@ }, "contextWindow": 2000000, "maxTokens": 2000000, + "compat": { + "reasoningEffortMap": { + "minimal": "low" + }, + "includeEncryptedReasoning": false, + "filterReasoningHistory": true, + "supportsImageDetailOriginal": false, + "omitReasoningEffort": false, + "supportsReasoningEffort": true + }, "thinking": { "mode": "effort", "efforts": [ @@ -98945,17 +102070,6 @@ "effortMap": { "minimal": "low" } - }, - "supportsComputerUse": false, - "compat": { - "reasoningEffortMap": { - "minimal": "low" - }, - "includeEncryptedReasoning": false, - "filterReasoningHistory": true, - "supportsImageDetailOriginal": false, - "omitReasoningEffort": false, - "supportsReasoningEffort": true } }, "grok-4.3": { @@ -98977,6 +102091,16 @@ }, "contextWindow": 1000000, "maxTokens": 1000000, + "compat": { + "reasoningEffortMap": { + "minimal": "low" + }, + "includeEncryptedReasoning": false, + "filterReasoningHistory": true, + "supportsImageDetailOriginal": false, + "omitReasoningEffort": false, + "supportsReasoningEffort": true + }, "thinking": { "mode": "effort", "efforts": [ @@ -98989,17 +102113,6 @@ "effortMap": { "minimal": "low" } - }, - "supportsComputerUse": false, - "compat": { - "reasoningEffortMap": { - "minimal": "low" - }, - "includeEncryptedReasoning": false, - "filterReasoningHistory": true, - "supportsImageDetailOriginal": false, - "omitReasoningEffort": false, - "supportsReasoningEffort": true } }, "grok-4.5": { @@ -99021,6 +102134,16 @@ }, "contextWindow": 500000, "maxTokens": 500000, + "compat": { + "reasoningEffortMap": { + "minimal": "low" + }, + "includeEncryptedReasoning": false, + "filterReasoningHistory": true, + "supportsImageDetailOriginal": false, + "omitReasoningEffort": false, + "supportsReasoningEffort": true + }, "thinking": { "mode": "effort", "efforts": [ @@ -99033,17 +102156,6 @@ "effortMap": { "minimal": "low" } - }, - "supportsComputerUse": false, - "compat": { - "reasoningEffortMap": { - "minimal": "low" - }, - "includeEncryptedReasoning": false, - "filterReasoningHistory": true, - "supportsImageDetailOriginal": false, - "omitReasoningEffort": false, - "supportsReasoningEffort": true } }, "grok-build": { @@ -99065,7 +102177,6 @@ }, "contextWindow": 512000, "maxTokens": 512000, - "supportsComputerUse": false, "compat": { "reasoningEffortMap": { "minimal": "low" @@ -99096,7 +102207,6 @@ }, "contextWindow": 256000, "maxTokens": 256000, - "supportsComputerUse": false, "compat": { "reasoningEffortMap": { "minimal": "low" @@ -99126,7 +102236,6 @@ }, "contextWindow": 200000, "maxTokens": 200000, - "supportsComputerUse": false, "compat": { "reasoningEffortMap": { "minimal": "low" @@ -100671,8 +103780,8 @@ "text" ], "cost": { - "input": 0.424593036, - "output": 1.910668662, + "input": 0.588, + "output": 2.646, "cacheRead": 0, "cacheWrite": 0 }, @@ -100887,10 +103996,10 @@ "image" ], "cost": { - "input": 0.422571632, - "output": 2.11285816, - "cacheRead": 0.08453343, - "cacheWrite": 0.001193975 + "input": 0.8848, + "output": 4.424, + "cacheRead": 0.177, + "cacheWrite": 0.0025 }, "contextWindow": 256000, "maxTokens": null, @@ -100918,10 +104027,10 @@ "image" ], "cost": { - "input": 0.422571632, - "output": 2.11285816, - "cacheRead": 0.08453343, - "cacheWrite": 0.00238795 + "input": 0.4424, + "output": 2.212, + "cacheRead": 0.0885, + "cacheWrite": 0.0025 }, "contextWindow": 256000, "maxTokens": null, @@ -101011,10 +104120,10 @@ "image" ], "cost": { - "input": 0.422571632, - "output": 2.11285816, - "cacheRead": 0.08453343, - "cacheWrite": 0.001193975 + "input": 0.8848, + "output": 4.424, + "cacheRead": 0.177, + "cacheWrite": 0.0025 }, "contextWindow": 256000, "maxTokens": null, @@ -101205,6 +104314,7 @@ "thinking": { "mode": "effort", "efforts": [ + "low", "high", "max" ] @@ -101231,6 +104341,7 @@ "thinking": { "mode": "effort", "efforts": [ + "low", "high", "max" ] @@ -101748,6 +104859,37 @@ "supportsComputerUse": false, "supportsComputerUseConfig": false }, + "google/gemma-4-26b-a4b-it": { + "id": "google/gemma-4-26b-a4b-it", + "name": "Gemma 4 26B A4B IT", + "api": "openai-completions", + "provider": "zenmux", + "baseUrl": "https://zenmux.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 0.15, + "output": 0.6, + "cacheRead": 0.015, + "cacheWrite": 0 + }, + "contextWindow": 262144, + "maxTokens": 16384, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + }, + "supportsComputerUse": false + }, "inclusionai/ling-1t": { "id": "inclusionai/ling-1t", "name": "Ling-1T", @@ -101778,9 +104920,9 @@ "text" ], "cost": { - "input": 0.1318155, - "output": 1.0984625, - "cacheRead": 0.0263631, + "input": 0.3, + "output": 2.5, + "cacheRead": 0.06, "cacheWrite": 0 }, "contextWindow": 262144, @@ -101789,7 +104931,7 @@ }, "inclusionai/ling-2.6-flash": { "id": "inclusionai/ling-2.6-flash", - "name": "Ling-2.6 Flash", + "name": "Ling-2.6-flash", "api": "openai-completions", "provider": "zenmux", "baseUrl": "https://zenmux.ai/api/v1", @@ -101813,19 +104955,29 @@ "api": "openai-completions", "provider": "zenmux", "baseUrl": "https://zenmux.ai/api/v1", - "reasoning": false, + "reasoning": true, "input": [ "text" ], "cost": { - "input": 0, - "output": 0, - "cacheRead": 0, + "input": 0.021, + "output": 0.063, + "cacheRead": 0.0042, "cacheWrite": 0 }, "contextWindow": 262144, "maxTokens": 32768, - "supportsComputerUse": false + "supportsComputerUse": false, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + } }, "inclusionai/ling-flash-2.0": { "id": "inclusionai/ling-flash-2.0", @@ -102063,9 +105215,9 @@ "text" ], "cost": { - "input": 0.135, - "output": 0.54, - "cacheRead": 0.027, + "input": 0.15, + "output": 0.6, + "cacheRead": 0.03, "cacheWrite": 0 }, "contextWindow": 256000, @@ -102144,9 +105296,9 @@ "text" ], "cost": { - "input": 0.444, - "output": 1.776, - "cacheRead": 0.09, + "input": 0.74, + "output": 2.96, + "cacheRead": 0.15, "cacheWrite": 0 }, "contextWindow": 256000, @@ -102155,7 +105307,7 @@ }, "meituan/longcat-2.0": { "id": "meituan/longcat-2.0", - "name": "LongCat-2.0", + "name": "LongCat 2.0", "api": "openai-completions", "provider": "zenmux", "baseUrl": "https://zenmux.ai/api/v1", @@ -102164,13 +105316,13 @@ "text" ], "cost": { - "input": 0.3, - "output": 1.18, - "cacheRead": 0.006, + "input": 0.75, + "output": 2.95, + "cacheRead": 0.015, "cacheWrite": 0 }, "contextWindow": 1000000, - "maxTokens": null, + "maxTokens": 262144, "thinking": { "mode": "effort", "efforts": [ @@ -102238,7 +105390,7 @@ "cost": { "input": 1.25, "output": 4.25, - "cacheRead": 0.125, + "cacheRead": 0.15, "cacheWrite": 0 }, "contextWindow": 1048576, @@ -102255,6 +105407,37 @@ }, "supportsComputerUse": false }, + "meta/muse-spark-1.2": { + "id": "meta/muse-spark-1.2", + "name": "Muse Spark 1.2", + "api": "openai-completions", + "provider": "zenmux", + "baseUrl": "https://zenmux.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 1.25, + "output": 4.25, + "cacheRead": 0.15, + "cacheWrite": 0 + }, + "contextWindow": 1048576, + "maxTokens": 1048576, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high", + "xhigh" + ] + }, + "supportsComputerUse": false + }, "minimax/minimax-m2": { "id": "minimax/minimax-m2", "name": "MiniMax M2", @@ -102475,7 +105658,7 @@ }, "mistralai/mistral-large-2512": { "id": "mistralai/mistral-large-2512", - "name": "Mistral Large 3 2512", + "name": "Mistral Large 3", "api": "openai-completions", "provider": "zenmux", "baseUrl": "https://zenmux.ai/api/v1", @@ -102716,7 +105899,7 @@ }, "moonshotai/kimi-k2.7-code-highspeed": { "id": "moonshotai/kimi-k2.7-code-highspeed", - "name": "Kimi K2.7 Code HighSpeed", + "name": "Kimi K2.7 Code High-Speed", "api": "openai-completions", "provider": "zenmux", "baseUrl": "https://zenmux.ai/api/v1", @@ -102855,7 +106038,7 @@ }, "openai/gpt-4.1-mini": { "id": "openai/gpt-4.1-mini", - "name": "GPT-4.1 Mini", + "name": "GPT-4.1 mini", "api": "openai-completions", "provider": "zenmux", "baseUrl": "https://zenmux.ai/api/v1", @@ -102876,7 +106059,7 @@ }, "openai/gpt-4.1-nano": { "id": "openai/gpt-4.1-nano", - "name": "GPT-4.1 Nano", + "name": "GPT-4.1 nano", "api": "openai-completions", "provider": "zenmux", "baseUrl": "https://zenmux.ai/api/v1", @@ -104072,10 +107255,10 @@ "image" ], "cost": { - "input": 0.134717, - "output": 0.808302, - "cacheRead": 0.0134717, - "cacheWrite": 0.16839625 + "input": 0.25, + "output": 1.5, + "cacheRead": 0.025, + "cacheWrite": 0.3125 }, "contextWindow": 1000000, "maxTokens": 65536, @@ -104149,7 +107332,7 @@ }, "qwen/qwen3.7-flash": { "id": "qwen/qwen3.7-flash", - "name": "Qwen3.7-Flash", + "name": "Qwen3.7 Flash", "api": "openai-completions", "provider": "zenmux", "baseUrl": "https://zenmux.ai/api/v1", @@ -104234,6 +107417,36 @@ ] } }, + "qwen/qwen3.8-max": { + "id": "qwen/qwen3.8-max", + "name": "Qwen3.8 Max", + "api": "openai-completions", + "provider": "zenmux", + "baseUrl": "https://zenmux.ai/api/v1", + "reasoning": true, + "input": [ + "text", + "image" + ], + "cost": { + "input": 2, + "output": 6, + "cacheRead": 0.17, + "cacheWrite": 2.5 + }, + "contextWindow": 1000000, + "maxTokens": 131072, + "thinking": { + "mode": "effort", + "efforts": [ + "minimal", + "low", + "medium", + "high" + ] + }, + "supportsComputerUse": false + }, "sapiens-ai/agnes-1.5-flash": { "id": "sapiens-ai/agnes-1.5-flash", "name": "Agnes-1.5-Flash", @@ -104520,7 +107733,7 @@ }, "tencent/hy3": { "id": "tencent/hy3", - "name": "Hy3", + "name": "Tencent Hy3", "api": "openai-completions", "provider": "zenmux", "baseUrl": "https://zenmux.ai/api/v1", @@ -104529,9 +107742,9 @@ "text" ], "cost": { - "input": 0.1323, - "output": 0.5301, - "cacheRead": 0.0333, + "input": 0.147, + "output": 0.589, + "cacheRead": 0.037, "cacheWrite": 0 }, "contextWindow": 262144, @@ -106096,4 +109309,4 @@ } } } -} +} \ No newline at end of file diff --git a/packages/catalog/src/provider-models/openai-compat.ts b/packages/catalog/src/provider-models/openai-compat.ts index d03c9fa29..26cdb52ed 100644 --- a/packages/catalog/src/provider-models/openai-compat.ts +++ b/packages/catalog/src/provider-models/openai-compat.ts @@ -2689,6 +2689,13 @@ const ALIBABA_TOKEN_PLAN_REASONING: ThinkingConfig = { mode: "effort", efforts: [Effort.Minimal, Effort.Low, Effort.Medium, Effort.High], }; +// Qwen3.8-Max uses the OpenAI `reasoning_effort` dialect; the generic Qwen +// dialect emits only the legacy binary `enable_thinking` toggle. +const ALIBABA_TOKEN_PLAN_QWEN_EFFORT_COMPAT: OpenAICompat = { + ...ALIBABA_TOKEN_PLAN_COMPAT, + supportsReasoningEffort: true, + thinkingFormat: "openai", +}; export const ALIBABA_TOKEN_PLAN_STATIC_MODELS: readonly ModelSpec<"openai-completions">[] = [ { @@ -2707,10 +2714,25 @@ export const ALIBABA_TOKEN_PLAN_STATIC_MODELS: readonly ModelSpec<"openai-comple efforts: [Effort.Low, Effort.High, Effort.XHigh], requiresEffort: true, }, - compat: { - ...ALIBABA_TOKEN_PLAN_COMPAT, - supportsReasoningEffort: true, + compat: ALIBABA_TOKEN_PLAN_QWEN_EFFORT_COMPAT, + }, + { + id: "qwen3.8-max", + name: "Qwen3.8 Max", + api: "openai-completions", + provider: "alibaba-token-plan", + baseUrl: ALIBABA_TOKEN_PLAN_BASE_URL, + reasoning: true, + input: ["text", "image"], + cost: ALIBABA_TOKEN_PLAN_COST, + contextWindow: 1_000_000, + maxTokens: 131_072, + thinking: { + mode: "effort", + efforts: [Effort.Low, Effort.Medium, Effort.XHigh], + defaultLevel: Effort.XHigh, }, + compat: ALIBABA_TOKEN_PLAN_QWEN_EFFORT_COMPAT, }, { id: "qwen3.7-max", diff --git a/packages/catalog/test/alibaba-token-plan.test.ts b/packages/catalog/test/alibaba-token-plan.test.ts index b4f539089..9f20d09a0 100644 --- a/packages/catalog/test/alibaba-token-plan.test.ts +++ b/packages/catalog/test/alibaba-token-plan.test.ts @@ -13,6 +13,7 @@ describe("QwenCloud Token Plan provider", () => { test("ships the documented Individual text-model allowlist", () => { expect(ALIBABA_TOKEN_PLAN_STATIC_MODELS.map(model => model.id)).toEqual([ "qwen3.8-max-preview", + "qwen3.8-max", "qwen3.7-max", "qwen3.7-plus", "qwen3.6-flash", @@ -143,6 +144,23 @@ describe("QwenCloud Token Plan provider", () => { contextWindow: 1_000_000, maxTokens: 64_000, }); + expect(models?.find(model => model.id === "qwen3.8-max")).toMatchObject({ + id: "qwen3.8-max", + provider: "alibaba-token-plan", + reasoning: true, + input: ["text", "image"], + contextWindow: 1_000_000, + maxTokens: 131_072, + thinking: { + mode: "effort", + efforts: [Effort.Low, Effort.Medium, Effort.XHigh], + defaultLevel: Effort.XHigh, + }, + compat: { + supportsReasoningEffort: true, + thinkingFormat: "openai", + }, + }); expect(options.dynamicModelsAuthoritative).toBe(true); }); diff --git a/packages/catalog/test/generated-policies.test.ts b/packages/catalog/test/generated-policies.test.ts index f8548f3dc..fbcb9a024 100644 --- a/packages/catalog/test/generated-policies.test.ts +++ b/packages/catalog/test/generated-policies.test.ts @@ -145,7 +145,7 @@ describe("generated model policies", () => { }); }); - it("preserves QwenCloud's mandatory qwen3.8 effort ladder", () => { + it("preserves QwenCloud's provider-authored qwen3.8 effort ladders", () => { const models: ModelSpec[] = [ createSpec({ id: "qwen3.8-max-preview", @@ -157,15 +157,32 @@ describe("generated model policies", () => { requiresEffort: true, }, }), + createSpec({ + id: "qwen3.8-max", + api: "openai-completions", + provider: "alibaba-token-plan", + thinking: { + mode: "effort", + efforts: [Effort.Low, Effort.Medium, Effort.XHigh], + defaultLevel: Effort.XHigh, + }, + }), ]; applyGeneratedModelPolicies(models); - expect(models[0]?.thinking).toEqual({ - mode: "effort", - efforts: [Effort.Low, Effort.High, Effort.XHigh], - requiresEffort: true, - }); + expect(models.map(model => model.thinking)).toEqual([ + { + mode: "effort", + efforts: [Effort.Low, Effort.High, Effort.XHigh], + requiresEffort: true, + }, + { + mode: "effort", + efforts: [Effort.Low, Effort.Medium, Effort.XHigh], + defaultLevel: Effort.XHigh, + }, + ]); }); it("pins zai glm-5.2 base id to 1M context", () => {