More fixes

Author: cte

src/api/providers/fetchers/__tests__/openrouter.spec.ts

@@ -49,7 +49,6 @@ describe("OpenRouter API", () => {
 				"agentica-org/deepcoder-14b-preview:free",
 				"aion-labs/aion-1.0",
 				"aion-labs/aion-1.0-mini",
-				"anthropic/claude-3.7-sonnet",
 				"anthropic/claude-3.7-sonnet:beta",
 				"anthropic/claude-3.7-sonnet:thinking",
 				"anthropic/claude-opus-4",
@@ -110,7 +109,6 @@ describe("OpenRouter API", () => {
 					.map(([id, _]) => id)
 					.sort(),
 			).toEqual([
-				"anthropic/claude-3.7-sonnet",
 				"anthropic/claude-3.7-sonnet:beta",
 				"anthropic/claude-3.7-sonnet:thinking",
 				"anthropic/claude-opus-4",
@@ -125,7 +123,7 @@ describe("OpenRouter API", () => {
 			).toEqual(["anthropic/claude-3.7-sonnet:thinking"])
 
 			expect(models["anthropic/claude-3.7-sonnet"]).toEqual({
-				maxTokens: 64000,
+				maxTokens: 8192,
 				contextWindow: 200000,
 				supportsImages: true,
 				supportsPromptCache: true,
@@ -135,9 +133,8 @@ describe("OpenRouter API", () => {
 				cacheReadsPrice: 0.3,
 				description: expect.any(String),
 				supportsComputerUse: true,
-				supportsReasoningBudget: true,
-				requiredReasoningBudget: false,
-				supportsReasoningEffort: true,
+				supportsReasoningBudget: false,
+				supportsReasoningEffort: false,
 				supportedParameters: ["max_tokens", "temperature", "reasoning", "include_reasoning"],
 			})
 
@@ -178,7 +175,7 @@ describe("OpenRouter API", () => {
 				{ id: "anthropic/claude-3.5-sonnet-20240620", maxTokens: 8192 },
 				{ id: "anthropic/claude-3.5-sonnet-20240620:beta", maxTokens: 8192 },
 				{ id: "anthropic/claude-3.5-sonnet:beta", maxTokens: 8192 },
-				{ id: "anthropic/claude-3.7-sonnet", maxTokens: 64000 },
+				{ id: "anthropic/claude-3.7-sonnet", maxTokens: 8192 },
 				{ id: "anthropic/claude-3.7-sonnet:beta", maxTokens: 128000 },
 				{ id: "anthropic/claude-3.7-sonnet:thinking", maxTokens: 128000 },
 			])
@@ -205,7 +202,6 @@ describe("OpenRouter API", () => {
 					description: undefined,
 					supportsReasoningBudget: false,
 					supportsReasoningEffort: undefined,
-					requiredReasoningBudget: false,
 					supportedParameters: undefined,
 				},
 				"Google AI Studio": {
@@ -220,7 +216,6 @@ describe("OpenRouter API", () => {
 					description: undefined,
 					supportsReasoningBudget: false,
 					supportsReasoningEffort: undefined,
-					requiredReasoningBudget: false,
 					supportedParameters: undefined,
 				},
 			})

src/api/providers/fetchers/openrouter.ts

@@ -197,7 +197,6 @@ export const parseOpenRouterModel = ({
 			id.startsWith("anthropic/claude-3.7") ||
 			id.startsWith("anthropic/claude-sonnet-4") ||
 			id.startsWith("anthropic/claude-opus-4"),
-		requiredReasoningBudget: id === "anthropic/claude-3.7-sonnet:thinking",
 		supportsReasoningEffort: supportedParameters ? supportedParameters.includes("reasoning") : undefined,
 		supportedParameters: supportedParameters ? supportedParameters.filter(isModelParameter) : undefined,
 	}
@@ -208,10 +207,19 @@ export const parseOpenRouterModel = ({
 		modelInfo.supportsComputerUse = true
 	}
 
-	// Special case: Claude 3.7 Sonnet Thinking supports up to 128k tokens if
-	// a beta flag is enabled (which we do).
+	// For backwards compatibility with the old model definitions we will
+	// continue to disable extending thinking for anthropic/claude-3.7-sonnet
+	// and force it for anthropic/claude-3.7-sonnet:thinking.
+
+	if (id === "anthropic/claude-3.7-sonnet") {
+		modelInfo.maxTokens = anthropicModels["claude-3-7-sonnet-20250219"].maxTokens
+		modelInfo.supportsReasoningBudget = false
+		modelInfo.supportsReasoningEffort = false
+	}
+
 	if (id === "anthropic/claude-3.7-sonnet:thinking") {
 		modelInfo.maxTokens = anthropicModels["claude-3-7-sonnet-20250219:thinking"].maxTokens
+		modelInfo.requiredReasoningBudget = true
 	}
 
 	return modelInfo

src/shared/api.ts

@@ -10,7 +10,7 @@ export type AnthropicModelId = keyof typeof anthropicModels
 export const anthropicDefaultModelId: AnthropicModelId = "claude-3-7-sonnet-20250219"
 export const anthropicModels = {
 	"claude-sonnet-4-20250514": {
-		maxTokens: 8192,
+		maxTokens: 64_000, // Overridden to 8k if `enableReasoningEffort` is false.
 		contextWindow: 200_000,
 		supportsImages: true,
 		supportsComputerUse: true,
@@ -22,7 +22,7 @@ export const anthropicModels = {
 		supportsReasoningBudget: true,
 	},
 	"claude-opus-4-20250514": {
-		maxTokens: 8192,
+		maxTokens: 32_000, // Overridden to 8k if `enableReasoningEffort` is false.
 		contextWindow: 200_000,
 		supportsImages: true,
 		supportsComputerUse: true,
@@ -34,7 +34,7 @@ export const anthropicModels = {
 		supportsReasoningBudget: true,
 	},
 	"claude-3-7-sonnet-20250219:thinking": {
-		maxTokens: 128_000,
+		maxTokens: 128_000, // Unlocked by passing `beta` flag to the model. Otherwise, it's 64k.
 		contextWindow: 200_000,
 		supportsImages: true,
 		supportsComputerUse: true,
@@ -47,7 +47,7 @@ export const anthropicModels = {
 		requiredReasoningBudget: true,
 	},
 	"claude-3-7-sonnet-20250219": {
-		maxTokens: 8192,
+		maxTokens: 8192, // Since we already have a `:thinking` virtual model we aren't setting `supportsReasoningBudget: true` here.
 		contextWindow: 200_000,
 		supportsImages: true,
 		supportsComputerUse: true,