Merge pull request #5251 from yzhg1983/fix/anthropic-opus-temperature

fix(ai): suppress deprecated temperature param for Claude Opus 4.7+

packages/ai/scripts/generate-models.ts

@@ -191,6 +191,11 @@ function isAnthropicAdaptiveThinkingModel(modelId: string): boolean {
 	);
 }
 
+function isAnthropicTemperatureUnsupportedModel(modelId: string): boolean {
+	const id = modelId.toLowerCase();
+	return id.includes("opus-4-7") || id.includes("opus-4.7") || id.includes("opus-4-8") || id.includes("opus-4.8");
+}
+
 function mergeAnthropicMessagesCompat(model: Model<Api>, compat: AnthropicMessagesCompat): void {
 	model.compat = { ...(model.compat as AnthropicMessagesCompat | undefined), ...compat };
 }
@@ -244,6 +249,9 @@ function applyThinkingLevelMetadata(model: Model<any>): void {
 	if (model.api === "anthropic-messages" && isAnthropicAdaptiveThinkingModel(model.id)) {
 		mergeAnthropicMessagesCompat(model, { forceAdaptiveThinking: true });
 	}
+	if (model.api === "anthropic-messages" && isAnthropicTemperatureUnsupportedModel(model.id)) {
+		mergeAnthropicMessagesCompat(model, { supportsTemperature: false });
+	}
 	if (model.api === "openai-completions" && model.id.includes("deepseek-v4")) {
 		mergeThinkingLevelMap(
 			model,

packages/ai/src/models.generated.ts

@@ -1850,7 +1850,7 @@ export const MODELS = {
 			api: "anthropic-messages",
 			provider: "anthropic",
 			baseUrl: "https://api.anthropic.com",
-			compat: {"forceAdaptiveThinking":true},
+			compat: {"forceAdaptiveThinking":true,"supportsTemperature":false},
 			reasoning: true,
 			thinkingLevelMap: {"xhigh":"xhigh"},
 			input: ["text", "image"],
@@ -1869,7 +1869,7 @@ export const MODELS = {
 			api: "anthropic-messages",
 			provider: "anthropic",
 			baseUrl: "https://api.anthropic.com",
-			compat: {"forceAdaptiveThinking":true},
+			compat: {"forceAdaptiveThinking":true,"supportsTemperature":false},
 			reasoning: true,
 			thinkingLevelMap: {"xhigh":"xhigh"},
 			input: ["text", "image"],
@@ -2958,7 +2958,26 @@ export const MODELS = {
 			api: "anthropic-messages",
 			provider: "cloudflare-ai-gateway",
 			baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/anthropic",
-			compat: {"forceAdaptiveThinking":true},
+			compat: {"forceAdaptiveThinking":true,"supportsTemperature":false},
+			reasoning: true,
+			thinkingLevelMap: {"xhigh":"xhigh"},
+			input: ["text", "image"],
+			cost: {
+				input: 5,
+				output: 25,
+				cacheRead: 0.5,
+				cacheWrite: 6.25,
+			},
+			contextWindow: 1000000,
+			maxTokens: 128000,
+		} satisfies Model<"anthropic-messages">,
+		"claude-opus-4-8": {
+			id: "claude-opus-4-8",
+			name: "Claude Opus 4.8",
+			api: "anthropic-messages",
+			provider: "cloudflare-ai-gateway",
+			baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/anthropic",
+			compat: {"forceAdaptiveThinking":true,"supportsTemperature":false},
 			reasoning: true,
 			thinkingLevelMap: {"xhigh":"xhigh"},
 			input: ["text", "image"],
@@ -3916,7 +3935,7 @@ export const MODELS = {
 			provider: "github-copilot",
 			baseUrl: "https://api.individual.githubcopilot.com",
 			headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
-			compat: {"forceAdaptiveThinking":true},
+			compat: {"forceAdaptiveThinking":true,"supportsTemperature":false},
 			reasoning: true,
 			thinkingLevelMap: {"xhigh":"xhigh"},
 			input: ["text", "image"],
@@ -3936,7 +3955,7 @@ export const MODELS = {
 			provider: "github-copilot",
 			baseUrl: "https://api.individual.githubcopilot.com",
 			headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
-			compat: {"forceAdaptiveThinking":true},
+			compat: {"forceAdaptiveThinking":true,"supportsTemperature":false},
 			reasoning: true,
 			thinkingLevelMap: {"xhigh":"xhigh"},
 			input: ["text", "image"],
@@ -7259,7 +7278,7 @@ export const MODELS = {
 			api: "anthropic-messages",
 			provider: "opencode",
 			baseUrl: "https://opencode.ai/zen",
-			compat: {"forceAdaptiveThinking":true},
+			compat: {"forceAdaptiveThinking":true,"supportsTemperature":false},
 			reasoning: true,
 			thinkingLevelMap: {"xhigh":"xhigh"},
 			input: ["text", "image"],
@@ -7278,7 +7297,7 @@ export const MODELS = {
 			api: "anthropic-messages",
 			provider: "opencode",
 			baseUrl: "https://opencode.ai/zen",
-			compat: {"forceAdaptiveThinking":true},
+			compat: {"forceAdaptiveThinking":true,"supportsTemperature":false},
 			reasoning: true,
 			thinkingLevelMap: {"xhigh":"xhigh"},
 			input: ["text", "image"],
@@ -8540,23 +8559,6 @@ export const MODELS = {
 			contextWindow: 2000000,
 			maxTokens: 30000,
 		} satisfies Model<"openai-completions">,
-		"baidu/ernie-4.5-21b-a3b": {
-			id: "baidu/ernie-4.5-21b-a3b",
-			name: "Baidu: ERNIE 4.5 21B A3B",
-			api: "openai-completions",
-			provider: "openrouter",
-			baseUrl: "https://openrouter.ai/api/v1",
-			reasoning: false,
-			input: ["text"],
-			cost: {
-				input: 0.07,
-				output: 0.28,
-				cacheRead: 0,
-				cacheWrite: 0,
-			},
-			contextWindow: 131072,
-			maxTokens: 8000,
-		} satisfies Model<"openai-completions">,
 		"baidu/ernie-4.5-vl-28b-a3b": {
 			id: "baidu/ernie-4.5-vl-28b-a3b",
 			name: "Baidu: ERNIE 4.5 VL 28B A3B",
@@ -9397,6 +9399,23 @@ export const MODELS = {
 			contextWindow: 131072,
 			maxTokens: 4096,
 		} satisfies Model<"openai-completions">,
+		"meta-llama/llama-4-maverick": {
+			id: "meta-llama/llama-4-maverick",
+			name: "Meta: Llama 4 Maverick",
+			api: "openai-completions",
+			provider: "openrouter",
+			baseUrl: "https://openrouter.ai/api/v1",
+			reasoning: false,
+			input: ["text", "image"],
+			cost: {
+				input: 0.15,
+				output: 0.6,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 1048576,
+			maxTokens: 16384,
+		} satisfies Model<"openai-completions">,
 		"meta-llama/llama-4-scout": {
 			id: "meta-llama/llama-4-scout",
 			name: "Meta: Llama 4 Scout",
@@ -9482,23 +9501,6 @@ export const MODELS = {
 			contextWindow: 204800,
 			maxTokens: 196608,
 		} satisfies Model<"openai-completions">,
-		"minimax/minimax-m2.5:free": {
-			id: "minimax/minimax-m2.5:free",
-			name: "MiniMax: MiniMax M2.5 (free)",
-			api: "openai-completions",
-			provider: "openrouter",
-			baseUrl: "https://openrouter.ai/api/v1",
-			reasoning: true,
-			input: ["text"],
-			cost: {
-				input: 0,
-				output: 0,
-				cacheRead: 0,
-				cacheWrite: 0,
-			},
-			contextWindow: 204800,
-			maxTokens: 8192,
-		} satisfies Model<"openai-completions">,
 		"minimax/minimax-m2.7": {
 			id: "minimax/minimax-m2.7",
 			name: "MiniMax: MiniMax M2.7",
@@ -9508,13 +9510,13 @@ export const MODELS = {
 			reasoning: true,
 			input: ["text"],
 			cost: {
-				input: 0.27899999999999997,
+				input: 0.26,
 				output: 1.2,
 				cacheRead: 0,
 				cacheWrite: 0,
 			},
 			contextWindow: 204800,
-			maxTokens: 131072,
+			maxTokens: 4096,
 		} satisfies Model<"openai-completions">,
 		"mistralai/codestral-2508": {
 			id: "mistralai/codestral-2508",
@@ -10916,13 +10918,13 @@ export const MODELS = {
 			reasoning: true,
 			input: ["text"],
 			cost: {
-				input: 0.03,
+				input: 0.029,
 				output: 0.14,
 				cacheRead: 0,
 				cacheWrite: 0,
 			},
 			contextWindow: 131072,
-			maxTokens: 131072,
+			maxTokens: 4096,
 		} satisfies Model<"openai-completions">,
 		"openai/gpt-oss-20b:free": {
 			id: "openai/gpt-oss-20b:free",
@@ -11230,23 +11232,6 @@ export const MODELS = {
 			contextWindow: 131072,
 			maxTokens: 16384,
 		} satisfies Model<"openai-completions">,
-		"qwen/qwen-2.5-7b-instruct": {
-			id: "qwen/qwen-2.5-7b-instruct",
-			name: "Qwen: Qwen2.5 7B Instruct",
-			api: "openai-completions",
-			provider: "openrouter",
-			baseUrl: "https://openrouter.ai/api/v1",
-			reasoning: false,
-			input: ["text"],
-			cost: {
-				input: 0.04,
-				output: 0.09999999999999999,
-				cacheRead: 0,
-				cacheWrite: 0,
-			},
-			contextWindow: 131072,
-			maxTokens: 32768,
-		} satisfies Model<"openai-completions">,
 		"qwen/qwen-plus": {
 			id: "qwen/qwen-plus",
 			name: "Qwen: Qwen-Plus",
@@ -11358,13 +11343,13 @@ export const MODELS = {
 			reasoning: true,
 			input: ["text"],
 			cost: {
-				input: 0.14950000000000002,
+				input: 0.09999999999999999,
-				output: 1.495,
+				output: 0.09999999999999999,
-				cacheRead: 0,
+				cacheRead: 0.09999999999999999,
 				cacheWrite: 0,
 			},
 			contextWindow: 262144,
-			maxTokens: 4096,
+			maxTokens: 262144,
 		} satisfies Model<"openai-completions">,
 		"qwen/qwen3-30b-a3b": {
 			id: "qwen/qwen3-30b-a3b",
@@ -11800,13 +11785,13 @@ export const MODELS = {
 			reasoning: true,
 			input: ["text", "image"],
 			cost: {
-				input: 0.13899999999999998,
+				input: 0.14,
 				output: 1,
-				cacheRead: 0,
+				cacheRead: 0.049999999999999996,
 				cacheWrite: 0,
 			},
 			contextWindow: 262144,
-			maxTokens: 4096,
+			maxTokens: 262144,
 		} satisfies Model<"openai-completions">,
 		"qwen/qwen3.5-397b-a17b": {
 			id: "qwen/qwen3.5-397b-a17b",
@@ -12148,23 +12133,6 @@ export const MODELS = {
 			contextWindow: 32768,
 			maxTokens: 32768,
 		} satisfies Model<"openai-completions">,
-		"upstage/solar-pro-3": {
-			id: "upstage/solar-pro-3",
-			name: "Upstage: Solar Pro 3",
-			api: "openai-completions",
-			provider: "openrouter",
-			baseUrl: "https://openrouter.ai/api/v1",
-			reasoning: true,
-			input: ["text"],
-			cost: {
-				input: 0.15,
-				output: 0.6,
-				cacheRead: 0.015,
-				cacheWrite: 0,
-			},
-			contextWindow: 128000,
-			maxTokens: 4096,
-		} satisfies Model<"openai-completions">,
 		"x-ai/grok-4.20": {
 			id: "x-ai/grok-4.20",
 			name: "xAI: Grok 4.20",
@@ -13451,7 +13419,7 @@ export const MODELS = {
 			api: "anthropic-messages",
 			provider: "vercel-ai-gateway",
 			baseUrl: "https://ai-gateway.vercel.sh",
-			compat: {"forceAdaptiveThinking":true},
+			compat: {"forceAdaptiveThinking":true,"supportsTemperature":false},
 			reasoning: true,
 			thinkingLevelMap: {"xhigh":"xhigh"},
 			input: ["text", "image"],
@@ -13470,7 +13438,7 @@ export const MODELS = {
 			api: "anthropic-messages",
 			provider: "vercel-ai-gateway",
 			baseUrl: "https://ai-gateway.vercel.sh",
-			compat: {"forceAdaptiveThinking":true},
+			compat: {"forceAdaptiveThinking":true,"supportsTemperature":false},
 			reasoning: true,
 			thinkingLevelMap: {"xhigh":"xhigh"},
 			input: ["text", "image"],
@@ -15264,6 +15232,23 @@ export const MODELS = {
 			contextWindow: 200000,
 			maxTokens: 8000,
 		} satisfies Model<"anthropic-messages">,
+		"stepfun/step-3.7-flash": {
+			id: "stepfun/step-3.7-flash",
+			name: "Step 3.7 Flash",
+			api: "anthropic-messages",
+			provider: "vercel-ai-gateway",
+			baseUrl: "https://ai-gateway.vercel.sh",
+			reasoning: true,
+			input: ["text", "image"],
+			cost: {
+				input: 0.19999999999999998,
+				output: 1.15,
+				cacheRead: 0.04,
+				cacheWrite: 0,
+			},
+			contextWindow: 256000,
+			maxTokens: 256000,
+		} satisfies Model<"anthropic-messages">,
 		"xai/grok-4.1-fast-non-reasoning": {
 			id: "xai/grok-4.1-fast-non-reasoning",
 			name: "Grok 4.1 Fast Non-Reasoning",

packages/ai/src/providers/anthropic.ts

@@ -176,6 +176,7 @@ function getAnthropicCompat(
 		sendSessionAffinityHeaders:
 			model.compat?.sendSessionAffinityHeaders ?? !!(isFireworks || isCloudflareAiGatewayAnthropic),
 		supportsCacheControlOnTools: model.compat?.supportsCacheControlOnTools ?? !isFireworks,
+		supportsTemperature: model.compat?.supportsTemperature ?? true,
 		allowEmptySignature: model.compat?.allowEmptySignature ?? false,
 	};
 }
@@ -896,15 +897,10 @@ function buildParams(
 	options?: AnthropicOptions,
 ): MessageCreateParamsStreaming {
 	const { cacheControl } = getCacheControl(model, options?.cacheRetention);
+	const compat = getAnthropicCompat(model);
 	const params: MessageCreateParamsStreaming = {
 		model: model.id,
-		messages: convertMessages(
+		messages: convertMessages(context.messages, model, isOAuthToken, cacheControl, compat.allowEmptySignature),
-			context.messages,
-			model,
-			isOAuthToken,
-			cacheControl,
-			getAnthropicCompat(model).allowEmptySignature,
-		),
 		max_tokens: options?.maxTokens ?? model.maxTokens,
 		stream: true,
 	};
@@ -936,13 +932,12 @@ function buildParams(
 		];
 	}
 
-	// Temperature is incompatible with extended thinking (adaptive or budget-based).
+	// Temperature is incompatible with extended thinking and unsupported on Claude Opus 4.7+.
-	if (options?.temperature !== undefined && !options?.thinkingEnabled) {
+	if (options?.temperature !== undefined && !options?.thinkingEnabled && compat.supportsTemperature) {
 		params.temperature = options.temperature;
 	}
 
 	if (context.tools && context.tools.length > 0) {
-		const compat = getAnthropicCompat(model);
 		params.tools = convertTools(
 			context.tools,
 			isOAuthToken,

packages/ai/src/types.ts

@@ -451,6 +451,12 @@ export interface AnthropicMessagesCompat {
 	 * Default: true.
 	 */
 	supportsCacheControlOnTools?: boolean;
+	/**
+	 * Whether the model accepts the Anthropic `temperature` request field.
+	 * Claude Opus 4.7+ rejects non-default temperature values.
+	 * Default: true.
+	 */
+	supportsTemperature?: boolean;
 	/**
 	 * Whether to force adaptive thinking (`thinking.type: "adaptive"` plus
 	 * `output_config.effort`) regardless of the model id. Built-in models that

packages/ai/test/anthropic-temperature-compat.test.ts

@@ -0,0 +1,104 @@
+import { describe, expect, it } from "vitest";
+import { getModel } from "../src/models.ts";
+import { streamSimple } from "../src/stream.ts";
+import type { Context, Model, SimpleStreamOptions } from "../src/types.ts";
+
+interface AnthropicTemperaturePayload {
+	temperature?: number;
+}
+
+class PayloadCaptured extends Error {
+	constructor() {
+		super("payload captured");
+		this.name = "PayloadCaptured";
+	}
+}
+
+function makeContext(): Context {
+	return {
+		messages: [{ role: "user", content: "Hello", timestamp: Date.now() }],
+	};
+}
+
+function makeCustomModel(compat?: Model<"anthropic-messages">["compat"]): Model<"anthropic-messages"> {
+	return {
+		id: "vendor--claude-opus-4-7",
+		name: "Vendor Proxy Opus 4.7",
+		api: "anthropic-messages",
+		provider: "vendor-proxy",
+		baseUrl: "http://127.0.0.1:9",
+		reasoning: true,
+		input: ["text"],
+		cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
+		contextWindow: 200000,
+		maxTokens: 32000,
+		compat,
+	};
+}
+
+async function capturePayload(
+	model: Model<"anthropic-messages">,
+	options?: SimpleStreamOptions,
+): Promise<AnthropicTemperaturePayload> {
+	let capturedPayload: AnthropicTemperaturePayload | undefined;
+
+	const payloadCaptureModel: Model<"anthropic-messages"> = {
+		...model,
+		baseUrl: "http://127.0.0.1:9",
+	};
+
+	const s = streamSimple(payloadCaptureModel, makeContext(), {
+		...options,
+		apiKey: "fake-key",
+		onPayload: (payload) => {
+			capturedPayload = payload as AnthropicTemperaturePayload;
+			throw new PayloadCaptured();
+		},
+	});
+
+	await s.result();
+
+	if (!capturedPayload) {
+		throw new Error("Expected payload to be captured before request failure");
+	}
+
+	return capturedPayload;
+}
+
+describe("Anthropic temperature compatibility", () => {
+	it("omits temperature for Claude Opus 4.7", async () => {
+		const payload = await capturePayload(getModel("anthropic", "claude-opus-4-7"), { temperature: 0 });
+
+		expect(payload.temperature).toBeUndefined();
+	});
+
+	it("omits temperature for Claude Opus 4.8", async () => {
+		const payload = await capturePayload(getModel("anthropic", "claude-opus-4-8"), { temperature: 0 });
+
+		expect(payload.temperature).toBeUndefined();
+	});
+
+	it("omits default temperature for Claude Opus 4.7", async () => {
+		const payload = await capturePayload(getModel("anthropic", "claude-opus-4-7"), { temperature: 1 });
+
+		expect(payload.temperature).toBeUndefined();
+	});
+
+	it("keeps temperature for Claude Opus 4.6", async () => {
+		const payload = await capturePayload(getModel("anthropic", "claude-opus-4-6"), { temperature: 0 });
+
+		expect(payload.temperature).toBe(0);
+	});
+
+	it("keeps temperature for Claude Sonnet 4.6", async () => {
+		const payload = await capturePayload(getModel("anthropic", "claude-sonnet-4-6"), { temperature: 0 });
+
+		expect(payload.temperature).toBe(0);
+	});
+
+	it("omits temperature for custom models with supportsTemperature disabled", async () => {
+		const payload = await capturePayload(makeCustomModel({ supportsTemperature: false }), { temperature: 0 });
+
+		expect(payload.temperature).toBeUndefined();
+	});
+});