test: update provider tests to expect capped maxTokens values

- Update tests to match the new behavior where modelMaxTokens is capped to model's actual limit

- Fix 4 failing tests in anthropic-vertex, anthropic, openrouter, and model-params specs

Fixes CI test failures

Author: MuriloFP

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

@@ -714,7 +714,7 @@ describe("VertexHandler", () => {
 			})
 
 			const result = handler.getModel()
-			expect(result.maxTokens).toBe(32_768)
+			expect(result.maxTokens).toBe(8192) // Capped to model's actual maxTokens
 			expect(result.reasoningBudget).toBeUndefined()
 			expect(result.temperature).toBe(0)
 		})

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

@@ -260,7 +260,7 @@ describe("AnthropicHandler", () => {
 			})
 
 			const result = handler.getModel()
-			expect(result.maxTokens).toBe(32_768)
+			expect(result.maxTokens).toBe(8192) // Capped to model's actual maxTokens
 			expect(result.reasoningBudget).toBeUndefined()
 			expect(result.temperature).toBe(0)
 		})

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

@@ -98,7 +98,7 @@ describe("OpenRouterHandler", () => {
 			})
 
 			const result = await handler.fetchModel()
-			expect(result.maxTokens).toBe(32_768) // Should use custom maxTokens
+			expect(result.maxTokens).toBe(32_768) // Not capped since model supports 128000
 			expect(result.reasoningBudget).toBeUndefined() // Use actual implementation value
 			expect(result.temperature).toBe(0) // Use actual implementation value
 		})
@@ -111,7 +111,7 @@ describe("OpenRouterHandler", () => {
 			})
 
 			const result = await handler.fetchModel()
-			expect(result.maxTokens).toBe(32_768)
+			expect(result.maxTokens).toBe(8192) // Capped to model's actual maxTokens
 			expect(result.reasoningBudget).toBeUndefined()
 			expect(result.temperature).toBe(0)
 		})

src/api/transform/__tests__/model-params.spec.ts

@@ -665,7 +665,7 @@ describe("getModelParams", () => {
 				model,
 			})
 
-			expect(result.maxTokens).toBe(20000)
+			expect(result.maxTokens).toBe(16000) // Capped to model's actual maxTokens
 			expect(result.reasoningBudget).toBe(10000)
 			expect(result.temperature).toBe(1.0) // Overridden for reasoning budget models
 			expect(result.reasoningEffort).toBeUndefined() // Budget takes precedence