feat: implement Claude Opus 5 support across all providers

Add full Claude Opus 5 support across all provider paths with correct
model definitions, adaptive-thinking handling, fetcher overrides, and
accurate pricing from Anthropic's published docs.

Model definitions (anthropic.ts, bedrock.ts, vertex.ts):
- 1M native context window (no beta header required)
- Adaptive-thinking binary toggle (supportsReasoningBudget/Binary)
- supportsTemperature: false (sampling params return a 400)
- Pricing: input $5.0, output $25.0, cacheWrites $6.25, cacheReads $0.50
  per million tokens

Provider handling:
- Add opus-5 to the Bedrock adaptive-thinking guard
- Add claude-opus-5 to OpenRouter and Vercel AI Gateway allow-lists
- Apply Opus 5 overrides (maxTokens, reasoning budget, temperature)
  in the openrouter, requesty, and vercel-ai-gateway fetchers

Tests:
- Add Opus 5 coverage across provider and fetcher test suites

Author: navedmerchant

packages/types/src/providers/anthropic.ts

@@ -145,6 +145,24 @@ export const anthropicModels = {
 		supportsReasoningBinary: true,
 		supportsTemperature: false,
 	},
+	"claude-opus-5": {
+		maxTokens: 128_000, // Overridden to 8k if `enableReasoningEffort` is false.
+		contextWindow: 1_000_000, // 1M context window native (no beta header required)
+		supportsImages: true,
+		supportsPromptCache: true,
+		inputPrice: 5.0, // $5 per million input tokens
+		outputPrice: 25.0, // $25 per million output tokens
+		cacheWritesPrice: 6.25, // $6.25 per million tokens
+		cacheReadsPrice: 0.5, // $0.50 per million tokens
+		// Opus 5 uses the same adaptive-thinking / binary-toggle convention as
+		// Opus 4.7+ and Sonnet 5 on the direct Anthropic provider path. Manual
+		// extended thinking (budget_tokens) is removed and returns a 400, and
+		// setting sampling parameters (temperature/top_p/top_k) returns a 400.
+		supportsReasoningBudget: true,
+		supportsReasoningBinary: true,
+		supportsTemperature: false,
+		description: "Claude Opus 5 is Anthropic's most capable model for complex agentic coding and enterprise work.",
+	},
 	"claude-fable-5": {
 		maxTokens: 128_000, // Overridden to 8k if `enableReasoningEffort` is false.
 		contextWindow: 1_000_000,

packages/types/src/providers/bedrock.ts

@@ -235,6 +235,23 @@ export const bedrockModels = {
 			},
 		],
 	},
+	"anthropic.claude-opus-5": {
+		maxTokens: 8192,
+		contextWindow: 1_000_000, // 1M context window native (no beta header required)
+		supportsImages: true,
+		supportsPromptCache: true,
+		supportsReasoningBudget: true,
+		supportsReasoningBinary: true,
+		supportsTemperature: false,
+		inputPrice: 5.0, // $5 per million input tokens
+		outputPrice: 25.0, // $25 per million output tokens
+		cacheWritesPrice: 6.25, // $6.25 per million tokens
+		cacheReadsPrice: 0.5, // $0.50 per million tokens
+		minTokensPerCachePoint: 1024,
+		maxCachePoints: 4,
+		cachableFields: ["system", "messages", "tools"],
+		description: "Claude Opus 5 is Anthropic's most capable model for complex agentic coding and enterprise work.",
+	},
 	"anthropic.claude-fable-5": {
 		maxTokens: 8192,
 		contextWindow: 1_000_000,
@@ -625,6 +642,7 @@ export const BEDROCK_1M_CONTEXT_MODEL_IDS = [
 // - Claude Opus 4.5
 // - Claude Opus 4.6
 // - Claude Opus 4.7
+// - Claude Opus 5
 // - Claude Fable 5 (cross-region inference only — can only be used through an inference profile)
 export const BEDROCK_GLOBAL_INFERENCE_MODEL_IDS = [
 	"anthropic.claude-sonnet-4-20250514-v1:0",
@@ -636,6 +654,7 @@ export const BEDROCK_GLOBAL_INFERENCE_MODEL_IDS = [
 	"anthropic.claude-opus-4-6-v1",
 	"anthropic.claude-opus-4-7",
 	"anthropic.claude-opus-4-8",
+	"anthropic.claude-opus-5",
 	"anthropic.claude-fable-5",
 ] as const
 

packages/types/src/providers/openrouter.ts

@@ -44,6 +44,7 @@ export const OPEN_ROUTER_PROMPT_CACHING_MODELS = new Set([
 	"anthropic/claude-opus-4.1",
 	"anthropic/claude-opus-4.5",
 	"anthropic/claude-opus-4.6",
+	"anthropic/claude-opus-5",
 	"anthropic/claude-fable-5",
 	"anthropic/claude-haiku-4.5",
 	"google/gemini-2.5-flash-preview",
@@ -76,6 +77,7 @@ export const OPEN_ROUTER_REASONING_BUDGET_MODELS = new Set([
 	"anthropic/claude-opus-4.1",
 	"anthropic/claude-opus-4.5",
 	"anthropic/claude-opus-4.6",
+	"anthropic/claude-opus-5",
 	"anthropic/claude-fable-5",
 	"anthropic/claude-sonnet-4",
 	"anthropic/claude-sonnet-4.5",

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

@@ -13,6 +13,7 @@ export const VERCEL_AI_GATEWAY_PROMPT_CACHING_MODELS = new Set([
 	"anthropic/claude-opus-4.1",
 	"anthropic/claude-opus-4.5",
 	"anthropic/claude-opus-4.6",
+	"anthropic/claude-opus-5",
 	"anthropic/claude-fable-5",
 	"anthropic/claude-sonnet-4",
 	"anthropic/claude-sonnet-4.6",
@@ -58,6 +59,7 @@ export const VERCEL_AI_GATEWAY_VISION_AND_TOOLS_MODELS = new Set([
 	"anthropic/claude-opus-4.1",
 	"anthropic/claude-opus-4.5",
 	"anthropic/claude-opus-4.6",
+	"anthropic/claude-opus-5",
 	"anthropic/claude-fable-5",
 	"anthropic/claude-sonnet-4",
 	"anthropic/claude-sonnet-4.5",

packages/types/src/providers/vertex.ts

@@ -462,6 +462,20 @@ export const vertexModels = {
 			},
 		],
 	},
+	"claude-opus-5": {
+		maxTokens: 8192,
+		contextWindow: 1_000_000, // 1M context window native (no beta header required)
+		supportsImages: true,
+		supportsPromptCache: true,
+		inputPrice: 5.0, // $5 per million input tokens
+		outputPrice: 25.0, // $25 per million output tokens
+		cacheWritesPrice: 6.25, // $6.25 per million tokens
+		cacheReadsPrice: 0.5, // $0.50 per million tokens
+		supportsReasoningBudget: true,
+		supportsReasoningBinary: true,
+		supportsTemperature: false,
+		description: "Claude Opus 5 is Anthropic's most capable model for complex agentic coding and enterprise work.",
+	},
 	"claude-fable-5": {
 		maxTokens: 8192,
 		contextWindow: 1_000_000,

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

@@ -1085,6 +1085,23 @@ describe("VertexHandler", () => {
 			expect(model.info.supportsTemperature).toBe(false)
 		})
 
+		it("should return Claude Opus 5 model info", () => {
+			const handler = new AnthropicVertexHandler({
+				apiModelId: "claude-opus-5",
+				vertexProjectId: "test-project",
+				vertexRegion: "us-central1",
+			})
+
+			const model = handler.getModel()
+			expect(model.id).toBe("claude-opus-5")
+			expect(model.info.maxTokens).toBe(8192)
+			expect(model.info.contextWindow).toBe(1_000_000)
+			expect(model.info.supportsReasoningBinary).toBe(true)
+			expect(model.info.supportsReasoningBudget).toBe(true)
+			expect(model.info.supportsPromptCache).toBe(true)
+			expect(model.info.supportsTemperature).toBe(false)
+		})
+
 		it("should not enable 1M context when flag is disabled", () => {
 			const handler = new AnthropicVertexHandler({
 				apiModelId: VERTEX_1M_CONTEXT_MODEL_IDS[0],
@@ -1392,6 +1409,35 @@ describe("VertexHandler", () => {
 			expect(request.thinking).not.toHaveProperty("budget_tokens")
 			expect(request.temperature).toBeUndefined()
 		})
+
+		it("should use adaptive thinking for Claude Opus 5", async () => {
+			const opusHandler = new AnthropicVertexHandler({
+				apiModelId: "claude-opus-5",
+				vertexProjectId: "test-project",
+				vertexRegion: "us-central1",
+				enableReasoningEffort: true,
+			})
+
+			const mockCreate = vitest.fn().mockImplementation(async () => ({
+				async *[Symbol.asyncIterator]() {
+					yield { type: "message_start", message: { usage: { input_tokens: 10, output_tokens: 5 } } }
+				},
+			}))
+			;(opusHandler["client"].messages as any).create = mockCreate
+
+			await opusHandler.createMessage("You are a helpful assistant", [{ role: "user", content: "Hello" }]).next()
+
+			expect(mockCreate).toHaveBeenCalledWith(
+				expect.objectContaining({
+					thinking: { type: "adaptive" },
+				}),
+				undefined,
+			)
+
+			const request = mockCreate.mock.calls[0][0]
+			expect(request.thinking).not.toHaveProperty("budget_tokens")
+			expect(request.temperature).toBeUndefined()
+		})
 	})
 
 	describe("native tool calling", () => {

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

@@ -456,6 +456,33 @@ describe("AnthropicHandler", () => {
 			expect(requestOptions?.headers?.["anthropic-beta"]).toContain("prompt-caching-2024-07-31")
 		})
 
+		it("should use adaptive thinking for Claude Opus 5 when reasoning is enabled", async () => {
+			const opusHandler = new AnthropicHandler({
+				apiKey: "test-api-key",
+				apiModelId: "claude-opus-5",
+				enableReasoningEffort: true,
+				modelMaxTokens: 32768,
+			})
+
+			const stream = opusHandler.createMessage(systemPrompt, [
+				{
+					role: "user",
+					content: [{ type: "text" as const, text: "Hello" }],
+				},
+			])
+
+			for await (const _chunk of stream) {
+				// Consume stream
+			}
+
+			const requestBody = mockCreate.mock.calls[mockCreate.mock.calls.length - 1]?.[0]
+			const requestOptions = mockCreate.mock.calls[mockCreate.mock.calls.length - 1]?.[1]
+			expect(requestBody?.thinking).toEqual({ type: "adaptive" })
+			expect(requestBody?.temperature).toBeUndefined()
+			expect(requestBody?.max_tokens).toBe(32768)
+			expect(requestOptions?.headers?.["anthropic-beta"]).toContain("prompt-caching-2024-07-31")
+		})
+
 		it("should send the custom model ID as-is and use adaptive thinking for a custom Sonnet-5-family model", async () => {
 			const customHandler = new AnthropicHandler({
 				apiKey: "test-api-key",
@@ -656,6 +683,23 @@ describe("AnthropicHandler", () => {
 			expect(model.reasoningBudget).toBeUndefined()
 		})
 
+		it("should handle Claude Opus 5 model correctly", () => {
+			const handler = new AnthropicHandler({
+				apiKey: "test-api-key",
+				apiModelId: "claude-opus-5",
+			})
+			const model = handler.getModel()
+			expect(model.id).toBe("claude-opus-5")
+			expect(model.info.maxTokens).toBe(128000)
+			expect(model.info.contextWindow).toBe(1000000)
+			expect(model.maxTokens).toBe(8192)
+			expect(model.info.supportsReasoningBinary).toBe(true)
+			expect(model.info.supportsReasoningBudget).toBe(true)
+			expect(model.info.supportsPromptCache).toBe(true)
+			expect(model.info.supportsTemperature).toBe(false)
+			expect(model.reasoningBudget).toBeUndefined()
+		})
+
 		it("should enable 1M context for Claude 4.5 Sonnet when beta flag is set", () => {
 			const handler = new AnthropicHandler({
 				apiKey: "test-api-key",

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

@@ -755,6 +755,37 @@ describe("AwsBedrockHandler", () => {
 			const model = handler.getModel()
 			expect(model.id).toBe("global.anthropic.claude-sonnet-5")
 		})
+
+		it("should return Claude Opus 5 model info", () => {
+			const handler = new AwsBedrockHandler({
+				apiModelId: "anthropic.claude-opus-5",
+				awsAccessKey: "test",
+				awsSecretKey: "test",
+				awsRegion: "us-east-1",
+			})
+
+			const model = handler.getModel()
+			expect(model.id).toBe("anthropic.claude-opus-5")
+			expect(model.info.contextWindow).toBe(1_000_000)
+			expect(model.info.supportsReasoningBinary).toBe(true)
+			expect(model.info.supportsReasoningBudget).toBe(true)
+			expect(model.info.supportsPromptCache).toBe(true)
+			expect(model.info.supportsTemperature).toBe(false)
+			expect(model.maxTokens).toBe(8192)
+		})
+
+		it("should apply global inference prefix for Claude Opus 5 when awsUseGlobalInference is true", () => {
+			const handler = new AwsBedrockHandler({
+				apiModelId: "anthropic.claude-opus-5",
+				awsAccessKey: "test",
+				awsSecretKey: "test",
+				awsRegion: "us-east-1",
+				awsUseGlobalInference: true,
+			})
+
+			const model = handler.getModel()
+			expect(model.id).toBe("global.anthropic.claude-opus-5")
+		})
 	})
 
 	describe("1M context beta feature", () => {
@@ -1487,6 +1518,36 @@ describe("AwsBedrockHandler", () => {
 			expect(commandArg.inferenceConfig?.temperature).toBeUndefined()
 		})
 
+		it("should send adaptive thinking with effort xhigh for Claude Opus 5 when reasoning is enabled", async () => {
+			// End-to-end regression guard for the Opus 5 handler branch. The
+			// isAdaptiveThinkingModel predicate is unit-covered, but a regression in
+			// the createMessage adaptive-thinking branch for this specific model
+			// wouldn't be caught without a request-level test.
+			const opus5Handler = new AwsBedrockHandler({
+				apiModelId: "anthropic.claude-opus-5",
+				awsAccessKey: "test-access-key",
+				awsSecretKey: "test-secret-key",
+				awsRegion: "us-east-1",
+				enableReasoningEffort: true,
+			})
+
+			const generator = opus5Handler.createMessage("System prompt", messages)
+			await generator.next()
+
+			expect(mockConverseStreamCommand).toHaveBeenCalled()
+			const commandArg = mockConverseStreamCommand.mock.calls[0][0] as any
+
+			// Opus 5 uses the same adaptive-thinking contract as Opus 4.7/4.8 —
+			// budget_tokens causes a 400, so thinking.type is "adaptive" with effort.
+			expect(commandArg.additionalModelRequestFields?.thinking).toEqual({
+				type: "adaptive",
+				display: "summarized",
+			})
+			expect(commandArg.additionalModelRequestFields?.output_config).toEqual({ effort: "xhigh" })
+			// Opus 5 rejects sampling parameters: temperature must be omitted entirely.
+			expect(commandArg.inferenceConfig?.temperature).toBeUndefined()
+		})
+
 		it("should omit thinking and temperature for Claude Opus 4.8 when reasoning is disabled", async () => {
 			const opus48Handler = new AwsBedrockHandler({
 				apiModelId: "anthropic.claude-opus-4-8",
@@ -1620,6 +1681,7 @@ describe("AwsBedrockHandler", () => {
 				expect(isAdaptiveThinkingModel("anthropic.claude-opus-4-8")).toBe(true)
 				expect(isAdaptiveThinkingModel("anthropic.claude-fable-5")).toBe(true)
 				expect(isAdaptiveThinkingModel("anthropic.claude-sonnet-5")).toBe(true)
+				expect(isAdaptiveThinkingModel("anthropic.claude-opus-5")).toBe(true)
 				// Future-proof Sonnet patterns — guarded even before a registry entry exists.
 				expect(isAdaptiveThinkingModel("anthropic.claude-sonnet-4-7")).toBe(true)
 				expect(isAdaptiveThinkingModel("anthropic.claude-sonnet-4-8")).toBe(true)
@@ -1629,12 +1691,14 @@ describe("AwsBedrockHandler", () => {
 				expect(isAdaptiveThinkingModel("us.anthropic.claude-opus-4-8")).toBe(true)
 				expect(isAdaptiveThinkingModel("global.anthropic.claude-fable-5")).toBe(true)
 				expect(isAdaptiveThinkingModel("global.anthropic.claude-sonnet-5")).toBe(true)
+				expect(isAdaptiveThinkingModel("global.anthropic.claude-opus-5")).toBe(true)
 				expect(isAdaptiveThinkingModel("eu.anthropic.claude-sonnet-4-7")).toBe(true)
 				expect(isAdaptiveThinkingModel("global.anthropic.claude-opus-4-8")).toBe(true)
 			})
 
 			it("returns false for older / non-adaptive models", () => {
 				expect(isAdaptiveThinkingModel("anthropic.claude-opus-4-6-v1")).toBe(false)
+				expect(isAdaptiveThinkingModel("anthropic.claude-opus-4-5-20251101-v1:0")).toBe(false)
 				expect(isAdaptiveThinkingModel("anthropic.claude-sonnet-4-6")).toBe(false)
 				expect(isAdaptiveThinkingModel("anthropic.claude-3-5-sonnet-20241022-v2:0")).toBe(false)
 				expect(isAdaptiveThinkingModel("amazon.nova-lite-v1:0")).toBe(false)