fix(ai): require explicit anthropic compat metadata

packages/ai/src/api/anthropic-messages.ts

@@ -170,16 +170,11 @@ const INTERLEAVED_THINKING_BETA = "interleaved-thinking-2025-05-14";
 function getAnthropicCompat(
 	model: Model<"anthropic-messages">,
 ): Required<Omit<AnthropicMessagesCompat, "forceAdaptiveThinking">> {
-	const isFireworks = model.provider === "fireworks";
-	const isCloudflareAiGatewayAnthropic =
-		model.provider === "cloudflare-ai-gateway" && model.baseUrl.includes("anthropic");
-
 	return {
-		supportsEagerToolInputStreaming: model.compat?.supportsEagerToolInputStreaming ?? !isFireworks,
+		supportsEagerToolInputStreaming: model.compat?.supportsEagerToolInputStreaming ?? true,
-		supportsLongCacheRetention: model.compat?.supportsLongCacheRetention ?? !isFireworks,
+		supportsLongCacheRetention: model.compat?.supportsLongCacheRetention ?? true,
-		sendSessionAffinityHeaders:
+		sendSessionAffinityHeaders: model.compat?.sendSessionAffinityHeaders ?? false,
-			model.compat?.sendSessionAffinityHeaders ?? (isFireworks || isCloudflareAiGatewayAnthropic),
+		supportsCacheControlOnTools: model.compat?.supportsCacheControlOnTools ?? true,
-		supportsCacheControlOnTools: model.compat?.supportsCacheControlOnTools ?? !isFireworks,
 		supportsTemperature: model.compat?.supportsTemperature ?? true,
 		allowEmptySignature: model.compat?.allowEmptySignature ?? false,
 	};

packages/ai/src/providers/openrouter.models.ts

@@ -909,7 +909,7 @@ export const OPENROUTER_MODELS = {
 			cacheRead: 0.2,
 			cacheWrite: 0.375,
 		},
-		contextWindow: 131072,
+		contextWindow: 65536,
 		maxTokens: 32768,
 	} satisfies Model<"openai-completions">,
 	"google/gemini-3.1-flash-lite": {
@@ -1108,7 +1108,7 @@ export const OPENROUTER_MODELS = {
 			cacheWrite: 0,
 		},
 		contextWindow: 262144,
-		maxTokens: 32768,
+		maxTokens: 8192,
 	} satisfies Model<"openai-completions">,
 	"ibm-granite/granite-4.1-8b": {
 		id: "ibm-granite/granite-4.1-8b",
@@ -1216,8 +1216,8 @@ export const OPENROUTER_MODELS = {
 			cacheRead: 0.06,
 			cacheWrite: 0,
 		},
-		contextWindow: 262144,
+		contextWindow: 256000,
-		maxTokens: 144000,
+		maxTokens: 80000,
 	} satisfies Model<"openai-completions">,
 	"liquid/lfm-2.5-1.2b-thinking:free": {
 		id: "liquid/lfm-2.5-1.2b-thinking:free",
@@ -1451,7 +1451,7 @@ export const OPENROUTER_MODELS = {
 			cacheWrite: 0,
 		},
 		contextWindow: 1048576,
-		maxTokens: 4096,
+		maxTokens: 512000,
 	} satisfies Model<"openai-completions">,
 	"mistralai/codestral-2508": {
 		id: "mistralai/codestral-2508",
@@ -1807,7 +1807,7 @@ export const OPENROUTER_MODELS = {
 		cost: {
 			input: 0.6,
 			output: 2.5,
-			cacheRead: 0.6,
+			cacheRead: 0,
 			cacheWrite: 0,
 		},
 		contextWindow: 262144,
@@ -2189,7 +2189,7 @@ export const OPENROUTER_MODELS = {
 			cacheWrite: 0,
 		},
 		contextWindow: 1047576,
-		maxTokens: 4096,
+		maxTokens: 32768,
 	} satisfies Model<"openai-completions">,
 	"openai/gpt-4.1-nano": {
 		id: "openai/gpt-4.1-nano",
@@ -2221,7 +2221,7 @@ export const OPENROUTER_MODELS = {
 		cost: {
 			input: 2.5,
 			output: 10,
-			cacheRead: 1.25,
+			cacheRead: 0,
 			cacheWrite: 0,
 		},
 		contextWindow: 128000,
@@ -2365,11 +2365,11 @@ export const OPENROUTER_MODELS = {
 		cost: {
 			input: 0.25,
 			output: 2,
-			cacheRead: 0.03,
+			cacheRead: 0.025,
 			cacheWrite: 0,
 		},
 		contextWindow: 400000,
-		maxTokens: 4096,
+		maxTokens: 128000,
 	} satisfies Model<"openai-completions">,
 	"openai/gpt-5-nano": {
 		id: "openai/gpt-5-nano",
@@ -2437,11 +2437,11 @@ export const OPENROUTER_MODELS = {
 		cost: {
 			input: 1.25,
 			output: 10,
-			cacheRead: 0.125,
+			cacheRead: 0.13,
 			cacheWrite: 0,
 		},
 		contextWindow: 128000,
-		maxTokens: 16384,
+		maxTokens: 32000,
 	} satisfies Model<"openai-completions">,
 	"openai/gpt-5.1-codex": {
 		id: "openai/gpt-5.1-codex",
@@ -2533,7 +2533,7 @@ export const OPENROUTER_MODELS = {
 			cacheWrite: 0,
 		},
 		contextWindow: 128000,
-		maxTokens: 32000,
+		maxTokens: 16384,
 	} satisfies Model<"openai-completions">,
 	"openai/gpt-5.2-codex": {
 		id: "openai/gpt-5.2-codex",
@@ -3799,11 +3799,11 @@ export const OPENROUTER_MODELS = {
 		cost: {
 			input: 0.14,
 			output: 1,
-			cacheRead: 0.05,
+			cacheRead: 0,
 			cacheWrite: 0,
 		},
 		contextWindow: 262144,
-		maxTokens: 81920,
+		maxTokens: 262144,
 	} satisfies Model<"openai-completions">,
 	"qwen/qwen3.5-397b-a17b": {
 		id: "qwen/qwen3.5-397b-a17b",
@@ -3838,8 +3838,8 @@ export const OPENROUTER_MODELS = {
 			cacheRead: 0,
 			cacheWrite: 0,
 		},
-		contextWindow: 256000,
+		contextWindow: 262144,
-		maxTokens: 32768,
+		maxTokens: 262144,
 	} satisfies Model<"openai-completions">,
 	"qwen/qwen3.5-flash-02-23": {
 		id: "qwen/qwen3.5-flash-02-23",
@@ -4429,11 +4429,11 @@ export const OPENROUTER_MODELS = {
 		cost: {
 			input: 0.98,
 			output: 3.08,
-			cacheRead: 0.182,
+			cacheRead: 0.49,
 			cacheWrite: 0,
 		},
 		contextWindow: 202752,
-		maxTokens: 4096,
+		maxTokens: 65535,
 	} satisfies Model<"openai-completions">,
 	"z-ai/glm-5.2": {
 		id: "z-ai/glm-5.2",

packages/ai/test/fireworks-models.test.ts

@@ -79,7 +79,16 @@ const tool: Tool = {
 	parameters: Type.Object({ value: Type.String() }),
 };
 
-function createFireworksModel(compat?: Model<"anthropic-messages">["compat"]): Model<"anthropic-messages"> {
+const FIREWORKS_ANTHROPIC_COMPAT = {
+	sendSessionAffinityHeaders: true,
+	supportsEagerToolInputStreaming: false,
+	supportsCacheControlOnTools: false,
+	supportsLongCacheRetention: false,
+} satisfies NonNullable<Model<"anthropic-messages">["compat"]>;
+
+function createFireworksModel(
+	compat: Model<"anthropic-messages">["compat"] = FIREWORKS_ANTHROPIC_COMPAT,
+): Model<"anthropic-messages"> {
 	return {
 		id: "accounts/fireworks/models/kimi-k2p6",
 		name: "Kimi K2.6",