fix(ai,coding-agent): replace deprecated kimi k2p5 model closes #3242

packages/ai/CHANGELOG.md

@@ -5,6 +5,7 @@
 ### Changed
 
 - Changed Anthropic prompt caching to add a `cache_control` breakpoint on the last tool definition, so tool schemas can be cached independently from transcript updates while preserving existing cache retention behavior ([#3260](https://github.com/badlogic/pi-mono/issues/3260))
+- Changed Kimi Coding model generation to normalize deprecated `k2p5` to `kimi-for-coding` from models.dev data and removed the old static fallback model list ([#3242](https://github.com/badlogic/pi-mono/issues/3242))
 
 ## [0.67.3] - 2026-04-15
 

packages/ai/scripts/generate-models.ts

@@ -606,13 +606,22 @@ async function loadModelsDevData(): Promise<Model<any>[]> {
 
 		// Process Kimi For Coding models
 		if (data["kimi-for-coding"]?.models) {
-			for (const [modelId, model] of Object.entries(data["kimi-for-coding"].models)) {
+			const kimiModels = data["kimi-for-coding"].models as Record<string, ModelsDevModel>;
+			const hasCanonicalModel = Object.prototype.hasOwnProperty.call(kimiModels, "kimi-for-coding");
+
+			for (const [modelId, model] of Object.entries(kimiModels)) {
 				const m = model as ModelsDevModel;
 				if (m.tool_call !== true) continue;
+				// models.dev still exposes deprecated "k2p5" in some snapshots.
+				// Normalize to the canonical model id and drop duplicates when canonical exists.
+				if (modelId === "k2p5" && hasCanonicalModel) continue;
+
+				const normalizedId = modelId === "k2p5" ? "kimi-for-coding" : modelId;
+				const normalizedName = modelId === "k2p5" ? "Kimi For Coding" : m.name || normalizedId;
 
 				models.push({
-					id: modelId,
-					name: m.name || modelId,
+					id: normalizedId,
+					name: normalizedName,
 					api: "anthropic-messages",
 					provider: "kimi-coding",
 					// Kimi For Coding's Anthropic-compatible API - SDK appends /v1/messages
@@ -1458,42 +1467,6 @@ async function generateModels() {
 	];
 	allModels.push(...vertexModels);
 
-	// Kimi For Coding models (Moonshot AI's Anthropic-compatible coding API)
-	// Static fallback in case models.dev doesn't have them yet
-	const KIMI_CODING_BASE_URL = "https://api.kimi.com/coding";
-	const kimiCodingModels: Model<"anthropic-messages">[] = [
-		{
-			id: "kimi-k2-thinking",
-			name: "Kimi K2 Thinking",
-			api: "anthropic-messages",
-			provider: "kimi-coding",
-			baseUrl: KIMI_CODING_BASE_URL,
-			reasoning: true,
-			input: ["text"],
-			cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
-			contextWindow: 262144,
-			maxTokens: 32768,
-		},
-		{
-			id: "k2p5",
-			name: "Kimi K2.5",
-			api: "anthropic-messages",
-			provider: "kimi-coding",
-			baseUrl: KIMI_CODING_BASE_URL,
-			reasoning: true,
-			input: ["text"],
-			cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
-			contextWindow: 262144,
-			maxTokens: 32768,
-		},
-	];
-	// Only add if not already present from models.dev
-	for (const model of kimiCodingModels) {
-		if (!allModels.some(m => m.provider === "kimi-coding" && m.id === model.id)) {
-			allModels.push(model);
-		}
-	}
-
 	const azureOpenAiModels: Model<Api>[] = allModels
 		.filter((model) => model.provider === "openai" && model.api === "openai-responses")
 		.map((model) => ({

packages/ai/src/models.generated.ts

@@ -4685,9 +4685,9 @@ export const MODELS = {
 		} satisfies Model<"openai-completions">,
 	},
 	"kimi-coding": {
-		"k2p5": {
-			id: "k2p5",
-			name: "Kimi K2.5",
+		"kimi-for-coding": {
+			id: "kimi-for-coding",
+			name: "Kimi For Coding",
 			api: "anthropic-messages",
 			provider: "kimi-coding",
 			baseUrl: "https://api.kimi.com/coding",
@@ -6618,6 +6618,40 @@ export const MODELS = {
 			contextWindow: 204800,
 			maxTokens: 128000,
 		} satisfies Model<"openai-completions">,
+		"qwen3.5-plus": {
+			id: "qwen3.5-plus",
+			name: "Qwen3.5 Plus",
+			api: "openai-completions",
+			provider: "opencode",
+			baseUrl: "https://opencode.ai/zen/v1",
+			reasoning: true,
+			input: ["text", "image"],
+			cost: {
+				input: 0.2,
+				output: 1.2,
+				cacheRead: 0.02,
+				cacheWrite: 0.25,
+			},
+			contextWindow: 262144,
+			maxTokens: 65536,
+		} satisfies Model<"openai-completions">,
+		"qwen3.6-plus": {
+			id: "qwen3.6-plus",
+			name: "Qwen3.6 Plus",
+			api: "openai-completions",
+			provider: "opencode",
+			baseUrl: "https://opencode.ai/zen/v1",
+			reasoning: true,
+			input: ["text", "image"],
+			cost: {
+				input: 0.5,
+				output: 3,
+				cacheRead: 0.05,
+				cacheWrite: 0.625,
+			},
+			contextWindow: 262144,
+			maxTokens: 65536,
+		} satisfies Model<"openai-completions">,
 	},
 	"opencode-go": {
 		"glm-5": {
@@ -6739,6 +6773,40 @@ export const MODELS = {
 			contextWindow: 204800,
 			maxTokens: 131072,
 		} satisfies Model<"anthropic-messages">,
+		"qwen3.5-plus": {
+			id: "qwen3.5-plus",
+			name: "Qwen3.5 Plus",
+			api: "openai-completions",
+			provider: "opencode-go",
+			baseUrl: "https://opencode.ai/zen/go/v1",
+			reasoning: true,
+			input: ["text", "image"],
+			cost: {
+				input: 0.2,
+				output: 1.2,
+				cacheRead: 0.02,
+				cacheWrite: 0.25,
+			},
+			contextWindow: 262144,
+			maxTokens: 65536,
+		} satisfies Model<"openai-completions">,
+		"qwen3.6-plus": {
+			id: "qwen3.6-plus",
+			name: "Qwen3.6 Plus",
+			api: "openai-completions",
+			provider: "opencode-go",
+			baseUrl: "https://opencode.ai/zen/go/v1",
+			reasoning: true,
+			input: ["text", "image"],
+			cost: {
+				input: 0.5,
+				output: 3,
+				cacheRead: 0.05,
+				cacheWrite: 0.625,
+			},
+			contextWindow: 262144,
+			maxTokens: 65536,
+		} satisfies Model<"openai-completions">,
 	},
 	"openrouter": {
 		"ai21/jamba-large-1.7": {

packages/ai/test/image-tool-result.test.ts

@@ -300,8 +300,8 @@ describe("Tool Results with Images", () => {
 		});
 	});
 
-	describe.skipIf(!process.env.KIMI_API_KEY)("Kimi For Coding Provider (k2p5)", () => {
-		const llm = getModel("kimi-coding", "k2p5");
+	describe.skipIf(!process.env.KIMI_API_KEY)("Kimi For Coding Provider (kimi-for-coding)", () => {
+		const llm = getModel("kimi-coding", "kimi-for-coding");
 
 		it("should handle tool result with only image", { retry: 3, timeout: 30000 }, async () => {
 			await handleToolWithImageResult(llm);

packages/ai/test/tokens.test.ts

@@ -192,7 +192,7 @@ describe("Token Statistics on Abort", () => {
 	});
 
 	describe.skipIf(!process.env.KIMI_API_KEY)("Kimi For Coding Provider", () => {
-		const llm = getModel("kimi-coding", "k2p5");
+		const llm = getModel("kimi-coding", "kimi-for-coding");
 
 		it("should include token stats when aborted mid-stream", { retry: 3, timeout: 30000 }, async () => {
 			await testTokensOnAbort(llm);

packages/coding-agent/CHANGELOG.md

@@ -9,6 +9,7 @@
 ### Fixed
 
 - Fixed flaky `edit-tool-no-full-redraw` TUI tests by waiting for asynchronous preview and preflight error rendering instead of relying on fixed render ticks.
+- Fixed `kimi-coding` default model selection to use `kimi-for-coding` instead of `kimi-k2-thinking` ([#3242](https://github.com/badlogic/pi-mono/issues/3242))
 
 ## [0.67.3] - 2026-04-15
 

packages/coding-agent/src/core/model-resolver.ts

@@ -34,7 +34,7 @@ export const defaultModelPerProvider: Record<KnownProvider, string> = {
 	huggingface: "moonshotai/Kimi-K2.5",
 	opencode: "claude-opus-4-6",
 	"opencode-go": "kimi-k2.5",
-	"kimi-coding": "kimi-k2-thinking",
+	"kimi-coding": "kimi-for-coding",
 };
 
 export interface ScopedModel {