fix: update provider tests to expect custom maxTokens for all models

- Fixed getModelMaxOutputTokens to always prioritize custom maxTokens
- Updated anthropic, anthropic-vertex, and openrouter tests to expect the new behavior
- Removed incorrect test that enforced old behavior for non-reasoning models
- All tests now pass with the corrected implementation

Author: MuriloFP

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

@@ -705,7 +705,7 @@ describe("VertexHandler", () => {
 			expect(result.temperature).toBe(1.0)
 		})
 
-		it("does not honor custom maxTokens for non-thinking models", () => {
+		it("honors custom maxTokens for all models", () => {
 			const handler = new AnthropicVertexHandler({
 				apiKey: "test-api-key",
 				apiModelId: "claude-3-7-sonnet@20250219",
@@ -714,7 +714,7 @@ describe("VertexHandler", () => {
 			})
 
 			const result = handler.getModel()
-			expect(result.maxTokens).toBe(8192)
+			expect(result.maxTokens).toBe(32_768)
 			expect(result.reasoningBudget).toBeUndefined()
 			expect(result.temperature).toBe(0)
 		})

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

@@ -251,7 +251,7 @@ describe("AnthropicHandler", () => {
 			expect(result.temperature).toBe(1.0)
 		})
 
-		it("does not honor custom maxTokens for non-thinking models", () => {
+		it("honors custom maxTokens for all models", () => {
 			const handler = new AnthropicHandler({
 				apiKey: "test-api-key",
 				apiModelId: "claude-3-7-sonnet-20250219",
@@ -260,7 +260,7 @@ describe("AnthropicHandler", () => {
 			})
 
 			const result = handler.getModel()
-			expect(result.maxTokens).toBe(8192)
+			expect(result.maxTokens).toBe(32_768)
 			expect(result.reasoningBudget).toBeUndefined()
 			expect(result.temperature).toBe(0)
 		})

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

@@ -89,7 +89,7 @@ describe("OpenRouterHandler", () => {
 			expect(result.info.supportsPromptCache).toBe(true)
 		})
 
-		it("honors custom maxTokens for thinking models", async () => {
+		it("honors custom maxTokens for all models", async () => {
 			const handler = new OpenRouterHandler({
 				openRouterApiKey: "test-key",
 				openRouterModelId: "anthropic/claude-3.7-sonnet:thinking",
@@ -98,20 +98,20 @@ describe("OpenRouterHandler", () => {
 			})
 
 			const result = await handler.fetchModel()
-			expect(result.maxTokens).toBe(128000) // Use actual implementation value
+			expect(result.maxTokens).toBe(32_768) // Should use custom maxTokens
 			expect(result.reasoningBudget).toBeUndefined() // Use actual implementation value
 			expect(result.temperature).toBe(0) // Use actual implementation value
 		})
 
-		it("does not honor custom maxTokens for non-thinking models", async () => {
+		it("honors custom maxTokens for non-thinking models", async () => {
 			const handler = new OpenRouterHandler({
 				...mockOptions,
 				modelMaxTokens: 32_768,
 				modelMaxThinkingTokens: 16_384,
 			})
 
 			const result = await handler.fetchModel()
-			expect(result.maxTokens).toBe(8192)
+			expect(result.maxTokens).toBe(32_768)
 			expect(result.reasoningBudget).toBeUndefined()
 			expect(result.temperature).toBe(0)
 		})

src/shared/api.ts

@@ -70,7 +70,7 @@ export const getModelMaxOutputTokens = ({
 		return settings.claudeCodeMaxOutputTokens || CLAUDE_CODE_DEFAULT_MAX_OUTPUT_TOKENS
 	}
 
-	// Check for user-configured modelMaxTokens FIRST (new logic)
+	// Check for user-configured modelMaxTokens
 	if (settings?.modelMaxTokens && settings.modelMaxTokens > 0) {
 		return settings.modelMaxTokens
 	}