feat: deprecate XML tool protocol selection, force native for new tasks

packages/types/src/providers/baseten.ts

@@ -33,6 +33,7 @@ export const basetenModels = {
 		contextWindow: 163_840,
 		supportsImages: false,
 		supportsPromptCache: false,
+		supportsNativeTools: true,
 		inputPrice: 2.55,
 		outputPrice: 5.95,
 		cacheWritesPrice: 0,
@@ -44,6 +45,7 @@ export const basetenModels = {
 		contextWindow: 163_840,
 		supportsImages: false,
 		supportsPromptCache: false,
+		supportsNativeTools: true,
 		inputPrice: 2.55,
 		outputPrice: 5.95,
 		cacheWritesPrice: 0,
@@ -55,6 +57,7 @@ export const basetenModels = {
 		contextWindow: 163_840,
 		supportsImages: false,
 		supportsPromptCache: false,
+		supportsNativeTools: true,
 		inputPrice: 0.77,
 		outputPrice: 0.77,
 		cacheWritesPrice: 0,
@@ -66,6 +69,7 @@ export const basetenModels = {
 		contextWindow: 163_840,
 		supportsImages: false,
 		supportsPromptCache: false,
+		supportsNativeTools: true,
 		inputPrice: 0.5,
 		outputPrice: 1.5,
 		cacheWritesPrice: 0,
@@ -86,11 +90,24 @@ export const basetenModels = {
 		description:
 			"DeepSeek's hybrid reasoning model with efficient long context scaling with GPT-5 level performance",
 	},
+	"openai/gpt-oss-120b": {
+		maxTokens: 16_384,
+		contextWindow: 128_072,
+		supportsImages: false,
+		supportsPromptCache: false,
+		supportsNativeTools: true,
+		inputPrice: 0.1,
+		outputPrice: 0.5,
+		cacheWritesPrice: 0,
+		cacheReadsPrice: 0,
+		description: "Extremely capable general-purpose LLM with strong, controllable reasoning capabilities",
+	},
 	"Qwen/Qwen3-235B-A22B-Instruct-2507": {
 		maxTokens: 16_384,
 		contextWindow: 262_144,
 		supportsImages: false,
 		supportsPromptCache: false,
+		supportsNativeTools: true,
 		inputPrice: 0.22,
 		outputPrice: 0.8,
 		cacheWritesPrice: 0,
@@ -102,24 +119,13 @@ export const basetenModels = {
 		contextWindow: 262_144,
 		supportsImages: false,
 		supportsPromptCache: false,
+		supportsNativeTools: true,
 		inputPrice: 0.38,
 		outputPrice: 1.53,
 		cacheWritesPrice: 0,
 		cacheReadsPrice: 0,
 		description: "Mixture-of-experts LLM with advanced coding and reasoning capabilities",
 	},
-	"openai/gpt-oss-120b": {
-		maxTokens: 16_384,
-		contextWindow: 128_072,
-		supportsImages: false,
-		supportsPromptCache: false,
-		supportsNativeTools: true,
-		inputPrice: 0.1,
-		outputPrice: 0.5,
-		cacheWritesPrice: 0,
-		cacheReadsPrice: 0,
-		description: "Extremely capable general-purpose LLM with strong, controllable reasoning capabilities",
-	},
 	"moonshotai/Kimi-K2-Instruct-0905": {
 		maxTokens: 16_384,
 		contextWindow: 262_000,

packages/types/src/providers/bedrock.ts

@@ -454,22 +454,6 @@ export const bedrockModels = {
 		outputPrice: 0.6,
 		description: "Amazon Titan Text Express",
 	},
-	"amazon.titan-text-embeddings-v1:0": {
-		maxTokens: 8192,
-		contextWindow: 8_000,
-		supportsImages: false,
-		supportsPromptCache: false,
-		inputPrice: 0.1,
-		description: "Amazon Titan Text Embeddings",
-	},
-	"amazon.titan-text-embeddings-v2:0": {
-		maxTokens: 8192,
-		contextWindow: 8_000,
-		supportsImages: false,
-		supportsPromptCache: false,
-		inputPrice: 0.02,
-		description: "Amazon Titan Text Embeddings V2",
-	},
 	"moonshot.kimi-k2-thinking": {
 		maxTokens: 32_000,
 		contextWindow: 262_144,

packages/types/src/providers/featherless.ts

@@ -13,6 +13,7 @@ export const featherlessModels = {
 		contextWindow: 32678,
 		supportsImages: false,
 		supportsPromptCache: false,
+		supportsNativeTools: true,
 		inputPrice: 0,
 		outputPrice: 0,
 		description: "DeepSeek V3 0324 model.",
@@ -22,6 +23,7 @@ export const featherlessModels = {
 		contextWindow: 32678,
 		supportsImages: false,
 		supportsPromptCache: false,
+		supportsNativeTools: true,
 		inputPrice: 0,
 		outputPrice: 0,
 		description: "DeepSeek R1 0528 model.",
@@ -41,6 +43,7 @@ export const featherlessModels = {
 		contextWindow: 32678,
 		supportsImages: false,
 		supportsPromptCache: false,
+		supportsNativeTools: true,
 		inputPrice: 0,
 		outputPrice: 0,
 		description: "GPT-OSS 120B model.",
@@ -57,4 +60,4 @@ export const featherlessModels = {
 	},
 } as const satisfies Record<string, ModelInfo>
 
-export const featherlessDefaultModelId: FeatherlessModelId = "deepseek-ai/DeepSeek-R1-0528"
+export const featherlessDefaultModelId: FeatherlessModelId = "moonshotai/Kimi-K2-Instruct"

packages/types/src/providers/groq.ts

@@ -5,12 +5,7 @@ export type GroqModelId =
 	| "llama-3.1-8b-instant"
 	| "llama-3.3-70b-versatile"
 	| "meta-llama/llama-4-scout-17b-16e-instruct"
-	| "meta-llama/llama-4-maverick-17b-128e-instruct"
-	| "mistral-saba-24b"
-	| "qwen-qwq-32b"
 	| "qwen/qwen3-32b"
-	| "deepseek-r1-distill-llama-70b"
-	| "moonshotai/kimi-k2-instruct"
 	| "moonshotai/kimi-k2-instruct-0905"
 	| "openai/gpt-oss-120b"
 	| "openai/gpt-oss-20b"
@@ -52,33 +47,6 @@ export const groqModels = {
 		outputPrice: 0.34,
 		description: "Meta Llama 4 Scout 17B Instruct model, 128K context.",
 	},
-	"meta-llama/llama-4-maverick-17b-128e-instruct": {
-		maxTokens: 8192,
-		contextWindow: 131072,
-		supportsImages: false,
-		supportsPromptCache: false,
-		inputPrice: 0.2,
-		outputPrice: 0.6,
-		description: "Meta Llama 4 Maverick 17B Instruct model, 128K context.",
-	},
-	"mistral-saba-24b": {
-		maxTokens: 8192,
-		contextWindow: 32768,
-		supportsImages: false,
-		supportsPromptCache: false,
-		inputPrice: 0.79,
-		outputPrice: 0.79,
-		description: "Mistral Saba 24B model, 32K context.",
-	},
-	"qwen-qwq-32b": {
-		maxTokens: 8192,
-		contextWindow: 131072,
-		supportsImages: false,
-		supportsPromptCache: false,
-		inputPrice: 0.29,
-		outputPrice: 0.39,
-		description: "Alibaba Qwen QwQ 32B model, 128K context.",
-	},
 	"qwen/qwen3-32b": {
 		maxTokens: 8192,
 		contextWindow: 131072,
@@ -90,25 +58,6 @@ export const groqModels = {
 		outputPrice: 0.59,
 		description: "Alibaba Qwen 3 32B model, 128K context.",
 	},
-	"deepseek-r1-distill-llama-70b": {
-		maxTokens: 8192,
-		contextWindow: 131072,
-		supportsImages: false,
-		supportsPromptCache: false,
-		inputPrice: 0.75,
-		outputPrice: 0.99,
-		description: "DeepSeek R1 Distill Llama 70B model, 128K context.",
-	},
-	"moonshotai/kimi-k2-instruct": {
-		maxTokens: 16384,
-		contextWindow: 131072,
-		supportsImages: false,
-		supportsPromptCache: true,
-		inputPrice: 1.0,
-		outputPrice: 3.0,
-		cacheReadsPrice: 0.5, // 50% discount for cached input tokens
-		description: "Moonshot AI Kimi K2 Instruct 1T model, 128K context.",
-	},
 	"moonshotai/kimi-k2-instruct-0905": {
 		maxTokens: 16384,
 		contextWindow: 262144,

packages/types/src/providers/sambanova.ts

@@ -7,9 +7,7 @@ export type SambaNovaModelId =
 	| "DeepSeek-R1"
 	| "DeepSeek-V3-0324"
 	| "DeepSeek-V3.1"
-	| "DeepSeek-R1-Distill-Llama-70B"
 	| "Llama-4-Maverick-17B-128E-Instruct"
-	| "Llama-3.3-Swallow-70B-Instruct-v0.4"
 	| "Qwen3-32B"
 	| "gpt-oss-120b"
 
@@ -72,15 +70,6 @@ export const sambaNovaModels = {
 		outputPrice: 4.5,
 		description: "DeepSeek V3.1 model with 32K context window.",
 	},
-	"DeepSeek-R1-Distill-Llama-70B": {
-		maxTokens: 8192,
-		contextWindow: 131072,
-		supportsImages: false,
-		supportsPromptCache: false,
-		inputPrice: 0.7,
-		outputPrice: 1.4,
-		description: "DeepSeek R1 distilled Llama 70B model with 128K context window.",
-	},
 	"Llama-4-Maverick-17B-128E-Instruct": {
 		maxTokens: 8192,
 		contextWindow: 131072,
@@ -92,15 +81,6 @@ export const sambaNovaModels = {
 		outputPrice: 1.8,
 		description: "Meta Llama 4 Maverick 17B 128E Instruct model with 128K context window.",
 	},
-	"Llama-3.3-Swallow-70B-Instruct-v0.4": {
-		maxTokens: 8192,
-		contextWindow: 16384,
-		supportsImages: false,
-		supportsPromptCache: false,
-		inputPrice: 0.6,
-		outputPrice: 1.2,
-		description: "Tokyotech Llama 3.3 Swallow 70B Instruct v0.4 model with 16K context window.",
-	},
 	"Qwen3-32B": {
 		maxTokens: 8192,
 		contextWindow: 8192,

packages/types/src/providers/vercel-ai-gateway.ts

@@ -90,6 +90,7 @@ export const vercelAiGatewayDefaultModelInfo: ModelInfo = {
 	contextWindow: 200000,
 	supportsImages: true,
 	supportsPromptCache: true,
+	supportsNativeTools: true,
 	inputPrice: 3,
 	outputPrice: 15,
 	cacheWritesPrice: 3.75,

packages/types/src/providers/vertex.ts

@@ -477,6 +477,7 @@ export const vertexModels = {
 		contextWindow: 131072,
 		supportsImages: false,
 		supportsPromptCache: false,
+		supportsNativeTools: true,
 		inputPrice: 0.35,
 		outputPrice: 1.15,
 		description: "Meta Llama 4 Maverick 17B Instruct model, 128K context.",
@@ -486,6 +487,7 @@ export const vertexModels = {
 		contextWindow: 163_840,
 		supportsImages: false,
 		supportsPromptCache: false,
+		supportsNativeTools: true,
 		inputPrice: 1.35,
 		outputPrice: 5.4,
 		description: "DeepSeek R1 (0528). Available in us-central1",
@@ -495,6 +497,7 @@ export const vertexModels = {
 		contextWindow: 163_840,
 		supportsImages: false,
 		supportsPromptCache: false,
+		supportsNativeTools: true,
 		inputPrice: 0.6,
 		outputPrice: 1.7,
 		description: "DeepSeek V3.1. Available in us-west2",
@@ -504,6 +507,7 @@ export const vertexModels = {
 		contextWindow: 131_072,
 		supportsImages: false,
 		supportsPromptCache: false,
+		supportsNativeTools: true,
 		inputPrice: 0.15,
 		outputPrice: 0.6,
 		description: "OpenAI gpt-oss 120B. Available in us-central1",
@@ -513,6 +517,7 @@ export const vertexModels = {
 		contextWindow: 131_072,
 		supportsImages: false,
 		supportsPromptCache: false,
+		supportsNativeTools: true,
 		inputPrice: 0.075,
 		outputPrice: 0.3,
 		description: "OpenAI gpt-oss 20B. Available in us-central1",
@@ -522,6 +527,7 @@ export const vertexModels = {
 		contextWindow: 262_144,
 		supportsImages: false,
 		supportsPromptCache: false,
+		supportsNativeTools: true,
 		inputPrice: 1.0,
 		outputPrice: 4.0,
 		description: "Qwen3 Coder 480B A35B Instruct. Available in us-south1",
@@ -531,6 +537,7 @@ export const vertexModels = {
 		contextWindow: 262_144,
 		supportsImages: false,
 		supportsPromptCache: false,
+		supportsNativeTools: true,
 		inputPrice: 0.25,
 		outputPrice: 1.0,
 		description: "Qwen3 235B A22B Instruct. Available in us-south1",

src/api/providers/__tests__/anthropic-vertex.spec.ts

@@ -1200,7 +1200,8 @@ describe("VertexHandler", () => {
 			)
 		})
 
-		it("should not include tools when toolProtocol is xml", async () => {
+		it("should include tools even when toolProtocol is set to xml (user preference now ignored)", async () => {
+			// XML protocol deprecation: user preference is now ignored when model supports native tools
 			handler = new AnthropicVertexHandler({
 				apiModelId: "claude-3-5-sonnet-v2@20241022",
 				vertexProjectId: "test-project",
@@ -1241,9 +1242,14 @@ describe("VertexHandler", () => {
 				// Just consume
 			}
 
+			// Native is forced when supportsNativeTools===true, so tools should still be included
 			expect(mockCreate).toHaveBeenCalledWith(
-				expect.not.objectContaining({
-					tools: expect.anything(),
+				expect.objectContaining({
+					tools: expect.arrayContaining([
+						expect.objectContaining({
+							name: "get_weather",
+						}),
+					]),
 				}),
 				undefined,
 			)

src/api/providers/__tests__/anthropic.spec.ts

@@ -451,8 +451,8 @@ describe("AnthropicHandler", () => {
 			)
 		})
 
-		it("should not include tools when toolProtocol is xml", async () => {
-			// Create handler with xml tool protocol in options
+		it("should include tools even when toolProtocol is set to xml (user preference now ignored)", async () => {
+			// XML protocol deprecation: user preference is now ignored when model supports native tools
 			const xmlHandler = new AnthropicHandler({
 				...mockOptions,
 				toolProtocol: "xml",
@@ -468,9 +468,14 @@ describe("AnthropicHandler", () => {
 				// Just consume
 			}
 
+			// Native is forced when supportsNativeTools===true, so tools should still be included
 			expect(mockCreate).toHaveBeenCalledWith(
-				expect.not.objectContaining({
-					tools: expect.anything(),
+				expect.objectContaining({
+					tools: expect.arrayContaining([
+						expect.objectContaining({
+							name: "get_weather",
+						}),
+					]),
 				}),
 				expect.anything(),
 			)