feat(coding-agent): improved model search ranking by match quality

- Prioritized fuzzy match quality over the most-recently-used model order when filtering the model browser.
- Added bucketing to match scores to ensure stable sorting when match quality is identical.
- Added unit tests to verify that exact query matches take precedence over the MRU model.

packages/coding-agent/CHANGELOG.md

@@ -21,6 +21,7 @@
 - Fixed `/btw` side-channel turns on Codex models such as `gpt-5.6-luna` by preserving the session websocket preference instead of forcing SSE, and made Esc dismiss the active `/btw` panel before interrupting loop/maintenance work. ([#5213](https://github.com/can1357/oh-my-pi/issues/5213))
 - Fixed the Model Hub role-assignment strip hiding the selected chip once the row overflowed; the strip now scrolls horizontally, truncating passed chips behind a leading ellipsis so the selection (plus one chip of lookahead) stays visible.
 - Fixed mouse hover and clicks in the /models Roles view landing one row above the pointer (the row mapping subtracted the status row twice).
+- Fixed model search keeping the most-recently-used model on top of the results: match quality now ranks first (an exact `gpt-5.5` beats the active `gpt-5.6-sol`), with MRU order only breaking ties between equally good matches.
 
 ## [16.4.5] - 2026-07-11
 

packages/coding-agent/src/modes/components/model-browser.ts

@@ -13,7 +13,7 @@ import { buildModel } from "@oh-my-pi/pi-catalog/build";
 import { modelsAreEqual } from "@oh-my-pi/pi-catalog/models";
 import {
 	type Component,
-	fuzzyFilter,
+	fuzzyRank,
 	Input,
 	matchesKey,
 	ScrollView,
@@ -432,10 +432,19 @@ export class ModelBrowser implements Component {
 		if (query.trim()) {
 			// Match against the displayed "provider/id" string so the user can
 			// type what they see: bare names, provider prefixes, or scoped
-			// queries all flow through the same fuzzy matcher. Skip role rank
-			// so a weakly matching default doesn't trump a stronger match.
-			const matches = fuzzyFilter(this.#baseItems, query, ({ provider, id }) => `${provider}/${id}`);
+			// queries all flow through the same fuzzy matcher.
+			const ranked = fuzzyRank(this.#baseItems, query, ({ provider, id }) => `${provider}/${id}`);
+			const matches = ranked.map(result => result.item);
+			// Match quality is the primary key while searching: an exact
+			// "gpt-5.5" must beat the MRU (or role-assigned) "gpt-5.6", so
+			// role rank is skipped and MRU only breaks ties. Scores are
+			// bucketed so sub-point position noise (provider-name length)
+			// can't split equally good matches; within a bucket the stable
+			// sort keeps sortModelItems' MRU/version order.
 			sortModelItems(matches, { roles: this.#roles, mruOrder: this.#mruOrder, skipRoleRank: true });
+			const buckets = new Map<ModelBrowserItem, number>();
+			for (const result of ranked) buckets.set(result.item, Math.round(result.score / 10));
+			matches.sort((a, b) => (buckets.get(a) ?? 0) - (buckets.get(b) ?? 0));
 			items = matches;
 		} else {
 			items = this.#baseItems;

packages/coding-agent/test/model-browser.test.ts

@@ -0,0 +1,64 @@
+import { describe, expect, test } from "bun:test";
+import type { Model } from "@oh-my-pi/pi-ai";
+import { buildModel } from "@oh-my-pi/pi-catalog/build";
+import { Settings } from "@oh-my-pi/pi-coding-agent/config/settings";
+import {
+	buildBrowserItems,
+	ModelBrowser,
+	sortModelItems,
+} from "@oh-my-pi/pi-coding-agent/modes/components/model-browser";
+
+function makeModel(provider: string, id: string): Model {
+	return buildModel({
+		id,
+		name: id,
+		api: "ollama-chat",
+		provider,
+		baseUrl: "https://example.com",
+		reasoning: false,
+		input: ["text"],
+		cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
+		contextWindow: 128_000,
+		maxTokens: 1024,
+	});
+}
+
+/** Browser preloaded with `models`, MRU-sorted like the hub does on sync. */
+function makeBrowser(models: Model[], mruOrder: string[]): ModelBrowser {
+	const browser = new ModelBrowser(Settings.isolated({}));
+	const items = buildBrowserItems(models);
+	sortModelItems(items, { mruOrder });
+	browser.setMruOrder(mruOrder);
+	browser.setItems(items);
+	return browser;
+}
+
+describe("ModelBrowser search ranking", () => {
+	test("an exact query match outranks the MRU model", () => {
+		// Regression: with gpt-5.6-sol as the active (MRU) model, typing
+		// "gpt-5.5" must select gpt-5.5, not keep the MRU pinned on top.
+		const browser = makeBrowser(
+			[
+				makeModel("openai-codex", "gpt-5.6-sol"),
+				makeModel("openai-codex", "gpt-5.6-luna"),
+				makeModel("openai-codex", "gpt-5.5"),
+				makeModel("openai-codex", "gpt-5.4"),
+			],
+			["openai-codex/gpt-5.6-sol", "openai-codex/gpt-5.6-luna"],
+		);
+
+		browser.setQuery("gpt-5.5");
+
+		expect(browser.getSelected()?.selector).toBe("openai-codex/gpt-5.5");
+	});
+
+	test("MRU breaks ties between equally good matches", () => {
+		// Same model id under two providers: match quality is identical, so
+		// the recently used provider must win over alphabetical order.
+		const browser = makeBrowser([makeModel("g0i", "gpt-5.5"), makeModel("zenmux", "gpt-5.5")], ["zenmux/gpt-5.5"]);
+
+		browser.setQuery("gpt-5.5");
+
+		expect(browser.getSelected()?.selector).toBe("zenmux/gpt-5.5");
+	});
+});