fix(catalog): kept qwen3.8 max preview on enable_thinking

Restored the preview to its main compat so the reasoning_effort dialect is scoped to qwen3.8-max, leaving the preview ladder unchanged.

packages/ai/test/openai-compat-policy.test.ts

@@ -175,4 +175,18 @@ describe("OpenAI compat policy", () => {
 			expect(params.chat_template_kwargs).toBeUndefined();
 		}
 	});
+
+	it("keeps Token Plan qwen3.8-max-preview on the enable_thinking dialect", () => {
+		// The preview rides Alibaba's binary enable_thinking toggle, not the
+		// OpenAI reasoning_effort control, so effort selections must not leak an
+		// unsupported reasoning_effort onto the wire.
+		const model = getBundledModel<"openai-completions">("alibaba-token-plan", "qwen3.8-max-preview");
+		const params = chatParams();
+		applyChatCompletionsCompatPolicy(
+			params,
+			resolveOpenAICompatPolicy(model, { endpoint: "chat-completions", reasoning: Effort.High }),
+		);
+		expect(params.enable_thinking).toBe(true);
+		expect(params.reasoning_effort).toBeUndefined();
+	});
 });

packages/catalog/src/models.json

@@ -8070,8 +8070,7 @@
 			},
 			"compat": {
 				"supportsDeveloperRole": false,
-				"supportsReasoningEffort": true,
-				"thinkingFormat": "openai"
+				"supportsReasoningEffort": true
 			}
 		}
 	},

packages/catalog/src/provider-models/openai-compat.ts

@@ -2714,7 +2714,10 @@ export const ALIBABA_TOKEN_PLAN_STATIC_MODELS: readonly ModelSpec<"openai-comple
 			efforts: [Effort.Low, Effort.High, Effort.XHigh],
 			requiresEffort: true,
 		},
-		compat: ALIBABA_TOKEN_PLAN_QWEN_EFFORT_COMPAT,
+		compat: {
+			...ALIBABA_TOKEN_PLAN_COMPAT,
+			supportsReasoningEffort: true,
+		},
 	},
 	{
 		id: "qwen3.8-max",