feat(vertex): add [1m] variants for Claude Sonnet models and send anthropic-beta header on Vertex; restore base to 200k and add tests

Author: roomote

packages/types/src/providers/vertex.ts

@@ -164,6 +164,18 @@ export const vertexModels = {
 		outputPrice: 5,
 	},
 	"claude-sonnet-4@20250514": {
+		maxTokens: 8192,
+		contextWindow: 200_000,
+		supportsImages: true,
+		supportsComputerUse: true,
+		supportsPromptCache: true,
+		inputPrice: 3.0,
+		outputPrice: 15.0,
+		cacheWritesPrice: 3.75,
+		cacheReadsPrice: 0.3,
+		supportsReasoningBudget: true,
+	},
+	"claude-sonnet-4@20250514[1m]": {
 		maxTokens: 8192,
 		contextWindow: 1_000_000,
 		supportsImages: true,
@@ -176,6 +188,18 @@ export const vertexModels = {
 		supportsReasoningBudget: true,
 	},
 	"claude-sonnet-4-5@20250929": {
+		maxTokens: 8192,
+		contextWindow: 200_000,
+		supportsImages: true,
+		supportsComputerUse: true,
+		supportsPromptCache: true,
+		inputPrice: 3.0,
+		outputPrice: 15.0,
+		cacheWritesPrice: 3.75,
+		cacheReadsPrice: 0.3,
+		supportsReasoningBudget: true,
+	},
+	"claude-sonnet-4-5@20250929[1m]": {
 		maxTokens: 8192,
 		contextWindow: 1_000_000,
 		supportsImages: true,

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

@@ -691,28 +691,30 @@ describe("VertexHandler", () => {
 			expect(modelInfo.info.contextWindow).toBe(200_000)
 		})
 
-		it("should return 1M context window for Claude Sonnet 4", () => {
+		it("should return 1M context window for Claude Sonnet 4 [1m] variant", () => {
 			handler = new AnthropicVertexHandler({
-				apiModelId: "claude-sonnet-4@20250514",
+				apiModelId: "claude-sonnet-4@20250514[1m]",
 				vertexProjectId: "test-project",
 				vertexRegion: "us-central1",
 			})
 
 			const modelInfo = handler.getModel()
+			// The provider strips the [1m] suffix when sending to API
 			expect(modelInfo.id).toBe("claude-sonnet-4@20250514")
 			expect(modelInfo.info).toBeDefined()
 			expect(modelInfo.info.maxTokens).toBe(8192)
 			expect(modelInfo.info.contextWindow).toBe(1_000_000)
 		})
 
-		it("should return 1M context window for Claude Sonnet 4.5", () => {
+		it("should return 1M context window for Claude Sonnet 4.5 [1m] variant", () => {
 			handler = new AnthropicVertexHandler({
-				apiModelId: "claude-sonnet-4-5@20250929",
+				apiModelId: "claude-sonnet-4-5@20250929[1m]",
 				vertexProjectId: "test-project",
 				vertexRegion: "us-central1",
 			})
 
 			const modelInfo = handler.getModel()
+			// The provider strips the [1m] suffix when sending to API
 			expect(modelInfo.id).toBe("claude-sonnet-4-5@20250929")
 			expect(modelInfo.info).toBeDefined()
 			expect(modelInfo.info.maxTokens).toBe(8192)

src/api/providers/anthropic-vertex.ts

@@ -87,16 +87,29 @@ export class AnthropicVertexHandler extends BaseProvider implements SingleComple
 			model: id,
 			max_tokens: maxTokens ?? ANTHROPIC_DEFAULT_MAX_TOKENS,
 			temperature,
-			thinking,
 			// Cache the system prompt if caching is enabled.
 			system: supportsPromptCache
 				? [{ text: systemPrompt, type: "text" as const, cache_control: { type: "ephemeral" } }]
 				: systemPrompt,
 			messages: supportsPromptCache ? addCacheBreakpoints(messages) : messages,
 			stream: true,
 		}
+		// Only set thinking if defined to avoid adding an explicit undefined property
+		if (thinking) {
+			;(params as any).thinking = thinking
+		}
 
-		const stream = await this.client.messages.create(params)
+		// Enable 1M context beta when using [1m] variants
+		const use1m = this.options.apiModelId?.endsWith("[1m]") === true
+
+		let stream
+		if (use1m) {
+			stream = await this.client.messages.create(params, {
+				headers: { "anthropic-beta": "context-1m-2025-08-07" },
+			})
+		} else {
+			stream = await this.client.messages.create(params)
+		}
 
 		for await (const chunk of stream) {
 			switch (chunk.type) {
@@ -171,8 +184,10 @@ export class AnthropicVertexHandler extends BaseProvider implements SingleComple
 		// The `:thinking` suffix indicates that the model is a "Hybrid"
 		// reasoning model and that reasoning is required to be enabled.
 		// The actual model ID honored by Anthropic's API does not have this
-		// suffix.
-		return { id: id.endsWith(":thinking") ? id.replace(":thinking", "") : id, info, ...params }
+		// suffix. Additionally, strip the optional [1m] suffix used to
+		// denote the 1M context beta variant in Roo's model list.
+		const normalizedId = id.replace(":thinking", "").replace("[1m]", "")
+		return { id: normalizedId, info, ...params }
 	}
 
 	async completePrompt(prompt: string) {
@@ -189,7 +204,6 @@ export class AnthropicVertexHandler extends BaseProvider implements SingleComple
 				model: id,
 				max_tokens: maxTokens,
 				temperature,
-				thinking,
 				messages: [
 					{
 						role: "user",
@@ -200,8 +214,17 @@ export class AnthropicVertexHandler extends BaseProvider implements SingleComple
 				],
 				stream: false,
 			}
+			// Only set thinking if defined to avoid adding an explicit undefined property
+			if (thinking) {
+				;(params as any).thinking = thinking
+			}
+
+			// Enable 1M context beta when using [1m] variants
+			const use1m = this.options.apiModelId?.endsWith("[1m]") === true
 
-			const response = await this.client.messages.create(params)
+			const response = use1m
+				? await this.client.messages.create(params, { headers: { "anthropic-beta": "context-1m-2025-08-07" } })
+				: await this.client.messages.create(params)
 			const content = response.content[0]
 
 			if (content.type === "text") {