Merge pull request #465 from router-for-me/think

Move thinking budget normalization from translators to executor

internal/api/modules/amp/routes.go

@@ -156,6 +156,7 @@ func (m *AmpModule) registerManagementRoutes(engine *gin.Engine, baseHandler *ha
 	rootMiddleware := []gin.HandlerFunc{m.managementAvailabilityMiddleware(), noCORSMiddleware(), m.localhostOnlyMiddleware()}
 	engine.GET("/threads/*path", append(rootMiddleware, proxyHandler)...)
 	engine.GET("/threads.rss", append(rootMiddleware, proxyHandler)...)
+	engine.GET("/news.rss", append(rootMiddleware, proxyHandler)...)
 
 	// Root-level auth routes for CLI login flow
 	// Amp uses multiple auth routes: /auth/cli-login, /auth/callback, /auth/sign-in, /auth/logout

internal/runtime/executor/aistudio_executor.go

@@ -309,11 +309,9 @@ func (e *AIStudioExecutor) translateRequest(req cliproxyexecutor.Request, opts c
 	to := sdktranslator.FromString("gemini")
 	payload := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), stream)
 	payload = applyThinkingMetadata(payload, req.Metadata, req.Model)
+	payload = util.ApplyDefaultThinkingIfNeeded(req.Model, payload)
 	payload = util.ConvertThinkingLevelToBudget(payload)
-	if budget := gjson.GetBytes(payload, "generationConfig.thinkingConfig.thinkingBudget"); budget.Exists() {
-		normalized := util.NormalizeThinkingBudget(req.Model, int(budget.Int()))
-		payload, _ = sjson.SetBytes(payload, "generationConfig.thinkingConfig.thinkingBudget", normalized)
-	}
+	payload = util.NormalizeGeminiThinkingBudget(req.Model, payload)
 	payload = util.StripThinkingConfigIfUnsupported(req.Model, payload)
 	payload = fixGeminiImageAspectRatio(req.Model, payload)
 	payload = applyPayloadConfig(e.cfg, req.Model, payload)

internal/runtime/executor/antigravity_executor.go

@@ -77,6 +77,7 @@ func (e *AntigravityExecutor) Execute(ctx context.Context, auth *cliproxyauth.Au
 	translated := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
 
 	translated = applyThinkingMetadataCLI(translated, req.Metadata, req.Model)
+	translated = util.ApplyDefaultThinkingIfNeededCLI(req.Model, translated)
 	translated = normalizeAntigravityThinking(req.Model, translated)
 
 	baseURLs := antigravityBaseURLFallbackOrder(auth)
@@ -171,6 +172,7 @@ func (e *AntigravityExecutor) ExecuteStream(ctx context.Context, auth *cliproxya
 	translated := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
 
 	translated = applyThinkingMetadataCLI(translated, req.Metadata, req.Model)
+	translated = util.ApplyDefaultThinkingIfNeededCLI(req.Model, translated)
 	translated = normalizeAntigravityThinking(req.Model, translated)
 
 	baseURLs := antigravityBaseURLFallbackOrder(auth)
@@ -839,9 +841,12 @@ func normalizeAntigravityThinking(model string, payload []byte) []byte {
 		effectiveMax, setDefaultMax := antigravityEffectiveMaxTokens(model, payload)
 		if effectiveMax > 0 && normalized >= effectiveMax {
 			normalized = effectiveMax - 1
-			if normalized < 1 {
-				normalized = 1
-			}
+		}
+		minBudget := antigravityMinThinkingBudget(model)
+		if minBudget > 0 && normalized >= 0 && normalized < minBudget {
+			// Budget is below minimum, remove thinking config entirely
+			payload, _ = sjson.DeleteBytes(payload, "request.generationConfig.thinkingConfig")
+			return payload
 		}
 		if setDefaultMax {
 			if res, errSet := sjson.SetBytes(payload, "request.generationConfig.maxOutputTokens", effectiveMax); errSet == nil {
@@ -869,3 +874,12 @@ func antigravityEffectiveMaxTokens(model string, payload []byte) (max int, fromM
 	}
 	return 0, false
 }
+
+// antigravityMinThinkingBudget returns the minimum thinking budget for a model.
+// Falls back to -1 if no model info is found.
+func antigravityMinThinkingBudget(model string) int {
+	if modelInfo := registry.GetGlobalRegistry().GetModelInfo(model); modelInfo != nil && modelInfo.Thinking != nil {
+		return modelInfo.Thinking.Min
+	}
+	return -1
+}

internal/runtime/executor/gemini_cli_executor.go

@@ -64,6 +64,8 @@ func (e *GeminiCLIExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth
 	to := sdktranslator.FromString("gemini-cli")
 	basePayload := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
 	basePayload = applyThinkingMetadataCLI(basePayload, req.Metadata, req.Model)
+	basePayload = util.ApplyDefaultThinkingIfNeededCLI(req.Model, basePayload)
+	basePayload = util.NormalizeGeminiCLIThinkingBudget(req.Model, basePayload)
 	basePayload = util.StripThinkingConfigIfUnsupported(req.Model, basePayload)
 	basePayload = fixGeminiCLIImageAspectRatio(req.Model, basePayload)
 	basePayload = applyPayloadConfigWithRoot(e.cfg, req.Model, "gemini", "request", basePayload)
@@ -199,6 +201,8 @@ func (e *GeminiCLIExecutor) ExecuteStream(ctx context.Context, auth *cliproxyaut
 	to := sdktranslator.FromString("gemini-cli")
 	basePayload := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
 	basePayload = applyThinkingMetadataCLI(basePayload, req.Metadata, req.Model)
+	basePayload = util.ApplyDefaultThinkingIfNeededCLI(req.Model, basePayload)
+	basePayload = util.NormalizeGeminiCLIThinkingBudget(req.Model, basePayload)
 	basePayload = util.StripThinkingConfigIfUnsupported(req.Model, basePayload)
 	basePayload = fixGeminiCLIImageAspectRatio(req.Model, basePayload)
 	basePayload = applyPayloadConfigWithRoot(e.cfg, req.Model, "gemini", "request", basePayload)

internal/runtime/executor/gemini_executor.go

@@ -80,6 +80,8 @@ func (e *GeminiExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, r
 	to := sdktranslator.FromString("gemini")
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
 	body = applyThinkingMetadata(body, req.Metadata, req.Model)
+	body = util.ApplyDefaultThinkingIfNeeded(req.Model, body)
+	body = util.NormalizeGeminiThinkingBudget(req.Model, body)
 	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
 	body = fixGeminiImageAspectRatio(req.Model, body)
 	body = applyPayloadConfig(e.cfg, req.Model, body)
@@ -169,6 +171,8 @@ func (e *GeminiExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
 	to := sdktranslator.FromString("gemini")
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
 	body = applyThinkingMetadata(body, req.Metadata, req.Model)
+	body = util.ApplyDefaultThinkingIfNeeded(req.Model, body)
+	body = util.NormalizeGeminiThinkingBudget(req.Model, body)
 	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
 	body = fixGeminiImageAspectRatio(req.Model, body)
 	body = applyPayloadConfig(e.cfg, req.Model, body)

internal/runtime/executor/gemini_vertex_executor.go

@@ -296,6 +296,8 @@ func (e *GeminiVertexExecutor) executeWithServiceAccount(ctx context.Context, au
 		}
 		body = util.ApplyGeminiThinkingConfig(body, budgetOverride, includeOverride)
 	}
+	body = util.ApplyDefaultThinkingIfNeeded(req.Model, body)
+	body = util.NormalizeGeminiThinkingBudget(req.Model, body)
 	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
 	body = fixGeminiImageAspectRatio(req.Model, body)
 	body = applyPayloadConfig(e.cfg, req.Model, body)
@@ -391,6 +393,8 @@ func (e *GeminiVertexExecutor) executeWithAPIKey(ctx context.Context, auth *clip
 		}
 		body = util.ApplyGeminiThinkingConfig(body, budgetOverride, includeOverride)
 	}
+	body = util.ApplyDefaultThinkingIfNeeded(req.Model, body)
+	body = util.NormalizeGeminiThinkingBudget(req.Model, body)
 	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
 	body = fixGeminiImageAspectRatio(req.Model, body)
 	body = applyPayloadConfig(e.cfg, req.Model, body)
@@ -487,6 +491,8 @@ func (e *GeminiVertexExecutor) executeStreamWithServiceAccount(ctx context.Conte
 		}
 		body = util.ApplyGeminiThinkingConfig(body, budgetOverride, includeOverride)
 	}
+	body = util.ApplyDefaultThinkingIfNeeded(req.Model, body)
+	body = util.NormalizeGeminiThinkingBudget(req.Model, body)
 	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
 	body = fixGeminiImageAspectRatio(req.Model, body)
 	body = applyPayloadConfig(e.cfg, req.Model, body)
@@ -599,6 +605,8 @@ func (e *GeminiVertexExecutor) executeStreamWithAPIKey(ctx context.Context, auth
 		}
 		body = util.ApplyGeminiThinkingConfig(body, budgetOverride, includeOverride)
 	}
+	body = util.ApplyDefaultThinkingIfNeeded(req.Model, body)
+	body = util.NormalizeGeminiThinkingBudget(req.Model, body)
 	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
 	body = fixGeminiImageAspectRatio(req.Model, body)
 	body = applyPayloadConfig(e.cfg, req.Model, body)

internal/translator/antigravity/claude/antigravity_claude_request.go

@@ -180,7 +180,6 @@ func ConvertClaudeRequestToAntigravity(modelName string, inputRawJSON []byte, _
 		if t.Get("type").String() == "enabled" {
 			if b := t.Get("budget_tokens"); b.Exists() && b.Type == gjson.Number {
 				budget := int(b.Int())
-				budget = util.NormalizeThinkingBudget(modelName, budget)
 				out, _ = sjson.Set(out, "request.generationConfig.thinkingConfig.thinkingBudget", budget)
 				out, _ = sjson.Set(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 			}

internal/translator/antigravity/openai/chat-completions/antigravity_openai_request.go

@@ -48,13 +48,13 @@ func ConvertOpenAIRequestToAntigravity(modelName string, inputRawJSON []byte, _
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", -1)
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 		case "low":
-			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 1024))
+			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", 1024)
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 		case "medium":
-			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 8192))
+			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", 8192)
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 		case "high":
-			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 32768))
+			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", 32768)
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 		default:
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", -1)
@@ -66,15 +66,15 @@ func ConvertOpenAIRequestToAntigravity(modelName string, inputRawJSON []byte, _
 	if !hasOfficialThinking && util.ModelSupportsThinking(modelName) {
 		if tc := gjson.GetBytes(rawJSON, "extra_body.google.thinking_config"); tc.Exists() && tc.IsObject() {
 			var setBudget bool
-			var normalized int
+			var budget int
 
 			if v := tc.Get("thinkingBudget"); v.Exists() {
-				normalized = util.NormalizeThinkingBudget(modelName, int(v.Int()))
-				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", normalized)
+				budget = int(v.Int())
+				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", budget)
 				setBudget = true
 			} else if v := tc.Get("thinking_budget"); v.Exists() {
-				normalized = util.NormalizeThinkingBudget(modelName, int(v.Int()))
-				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", normalized)
+				budget = int(v.Int())
+				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", budget)
 				setBudget = true
 			}
 
@@ -82,7 +82,7 @@ func ConvertOpenAIRequestToAntigravity(modelName string, inputRawJSON []byte, _
 				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", v.Bool())
 			} else if v := tc.Get("include_thoughts"); v.Exists() {
 				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", v.Bool())
-			} else if setBudget && normalized != 0 {
+			} else if setBudget && budget != 0 {
 				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 			}
 		}
@@ -94,7 +94,7 @@ func ConvertOpenAIRequestToAntigravity(modelName string, inputRawJSON []byte, _
 		if t := gjson.GetBytes(rawJSON, "thinking"); t.Exists() && t.IsObject() {
 			if t.Get("type").String() == "enabled" {
 				if b := t.Get("budget_tokens"); b.Exists() && b.Type == gjson.Number {
-					budget := util.NormalizeThinkingBudget(modelName, int(b.Int()))
+					budget := int(b.Int())
 					out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", budget)
 					out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 				}
@@ -102,15 +102,6 @@ func ConvertOpenAIRequestToAntigravity(modelName string, inputRawJSON []byte, _
 		}
 	}
 
-	// For gemini-3-pro-preview, always send default thinkingConfig when none specified.
-	// This matches the official Gemini CLI behavior which always sends:
-	// { thinkingBudget: -1, includeThoughts: true }
-	// See: ai-gemini-cli/packages/core/src/config/defaultModelConfigs.ts
-	if !gjson.GetBytes(out, "request.generationConfig.thinkingConfig").Exists() && modelName == "gemini-3-pro-preview" {
-		out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", -1)
-		out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
-	}
-
 	// Temperature/top_p/top_k/max_tokens
 	if tr := gjson.GetBytes(rawJSON, "temperature"); tr.Exists() && tr.Type == gjson.Number {
 		out, _ = sjson.SetBytes(out, "request.generationConfig.temperature", tr.Num)

internal/translator/gemini-cli/claude/gemini-cli_claude_request.go

@@ -165,7 +165,6 @@ func ConvertClaudeRequestToCLI(modelName string, inputRawJSON []byte, _ bool) []
 		if t.Get("type").String() == "enabled" {
 			if b := t.Get("budget_tokens"); b.Exists() && b.Type == gjson.Number {
 				budget := int(b.Int())
-				budget = util.NormalizeThinkingBudget(modelName, budget)
 				out, _ = sjson.Set(out, "request.generationConfig.thinkingConfig.thinkingBudget", budget)
 				out, _ = sjson.Set(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 			}

internal/translator/gemini-cli/openai/chat-completions/gemini-cli_openai_request.go

@@ -48,13 +48,13 @@ func ConvertOpenAIRequestToGeminiCLI(modelName string, inputRawJSON []byte, _ bo
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", -1)
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 		case "low":
-			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 1024))
+			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", 1024)
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 		case "medium":
-			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 8192))
+			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", 8192)
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 		case "high":
-			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 32768))
+			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", 32768)
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 		default:
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", -1)
@@ -66,15 +66,15 @@ func ConvertOpenAIRequestToGeminiCLI(modelName string, inputRawJSON []byte, _ bo
 	if !hasOfficialThinking && util.ModelSupportsThinking(modelName) {
 		if tc := gjson.GetBytes(rawJSON, "extra_body.google.thinking_config"); tc.Exists() && tc.IsObject() {
 			var setBudget bool
-			var normalized int
+			var budget int
 
 			if v := tc.Get("thinkingBudget"); v.Exists() {
-				normalized = util.NormalizeThinkingBudget(modelName, int(v.Int()))
-				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", normalized)
+				budget = int(v.Int())
+				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", budget)
 				setBudget = true
 			} else if v := tc.Get("thinking_budget"); v.Exists() {
-				normalized = util.NormalizeThinkingBudget(modelName, int(v.Int()))
-				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", normalized)
+				budget = int(v.Int())
+				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", budget)
 				setBudget = true
 			}
 
@@ -82,21 +82,12 @@ func ConvertOpenAIRequestToGeminiCLI(modelName string, inputRawJSON []byte, _ bo
 				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", v.Bool())
 			} else if v := tc.Get("include_thoughts"); v.Exists() {
 				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", v.Bool())
-			} else if setBudget && normalized != 0 {
+			} else if setBudget && budget != 0 {
 				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 			}
 		}
 	}
 
-	// For gemini-3-pro-preview, always send default thinkingConfig when none specified.
-	// This matches the official Gemini CLI behavior which always sends:
-	// { thinkingBudget: -1, includeThoughts: true }
-	// See: ai-gemini-cli/packages/core/src/config/defaultModelConfigs.ts
-	if !gjson.GetBytes(out, "request.generationConfig.thinkingConfig").Exists() && modelName == "gemini-3-pro-preview" {
-		out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", -1)
-		out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
-	}
-
 	// Temperature/top_p/top_k
 	if tr := gjson.GetBytes(rawJSON, "temperature"); tr.Exists() && tr.Type == gjson.Number {
 		out, _ = sjson.SetBytes(out, "request.generationConfig.temperature", tr.Num)

internal/translator/gemini/claude/gemini_claude_request.go

@@ -158,7 +158,6 @@ func ConvertClaudeRequestToGemini(modelName string, inputRawJSON []byte, _ bool)
 		if t.Get("type").String() == "enabled" {
 			if b := t.Get("budget_tokens"); b.Exists() && b.Type == gjson.Number {
 				budget := int(b.Int())
-				budget = util.NormalizeThinkingBudget(modelName, budget)
 				out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", budget)
 				out, _ = sjson.Set(out, "generationConfig.thinkingConfig.include_thoughts", true)
 			}

internal/translator/gemini/openai/chat-completions/gemini_openai_request.go

@@ -48,13 +48,13 @@ func ConvertOpenAIRequestToGemini(modelName string, inputRawJSON []byte, _ bool)
 			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", -1)
 			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.include_thoughts", true)
 		case "low":
-			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 1024))
+			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", 1024)
 			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.include_thoughts", true)
 		case "medium":
-			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 8192))
+			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", 8192)
 			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.include_thoughts", true)
 		case "high":
-			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 32768))
+			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", 32768)
 			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.include_thoughts", true)
 		default:
 			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", -1)
@@ -66,15 +66,15 @@ func ConvertOpenAIRequestToGemini(modelName string, inputRawJSON []byte, _ bool)
 	if !hasOfficialThinking && util.ModelSupportsThinking(modelName) {
 		if tc := gjson.GetBytes(rawJSON, "extra_body.google.thinking_config"); tc.Exists() && tc.IsObject() {
 			var setBudget bool
-			var normalized int
+			var budget int
 
 			if v := tc.Get("thinkingBudget"); v.Exists() {
-				normalized = util.NormalizeThinkingBudget(modelName, int(v.Int()))
-				out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", normalized)
+				budget = int(v.Int())
+				out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", budget)
 				setBudget = true
 			} else if v := tc.Get("thinking_budget"); v.Exists() {
-				normalized = util.NormalizeThinkingBudget(modelName, int(v.Int()))
-				out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", normalized)
+				budget = int(v.Int())
+				out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", budget)
 				setBudget = true
 			}
 
@@ -82,7 +82,7 @@ func ConvertOpenAIRequestToGemini(modelName string, inputRawJSON []byte, _ bool)
 				out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.include_thoughts", v.Bool())
 			} else if v := tc.Get("include_thoughts"); v.Exists() {
 				out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.include_thoughts", v.Bool())
-			} else if setBudget && normalized != 0 {
+			} else if setBudget && budget != 0 {
 				out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.include_thoughts", true)
 			}
 		}

internal/translator/gemini/openai/responses/gemini_openai-responses_request.go

@@ -400,16 +400,16 @@ func ConvertOpenAIResponsesRequestToGemini(modelName string, inputRawJSON []byte
 			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", -1)
 			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.include_thoughts", true)
 		case "minimal":
-			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 1024))
+			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", 1024)
 			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.include_thoughts", true)
 		case "low":
-			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 4096))
+			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", 4096)
 			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.include_thoughts", true)
 		case "medium":
-			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 8192))
+			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", 8192)
 			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.include_thoughts", true)
 		case "high":
-			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 32768))
+			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", 32768)
 			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.include_thoughts", true)
 		default:
 			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", -1)
@@ -421,32 +421,22 @@ func ConvertOpenAIResponsesRequestToGemini(modelName string, inputRawJSON []byte
 	if !hasOfficialThinking && util.ModelSupportsThinking(modelName) {
 		if tc := root.Get("extra_body.google.thinking_config"); tc.Exists() && tc.IsObject() {
 			var setBudget bool
-			var normalized int
+			var budget int
 			if v := tc.Get("thinking_budget"); v.Exists() {
-				normalized = util.NormalizeThinkingBudget(modelName, int(v.Int()))
-				out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", normalized)
+				budget = int(v.Int())
+				out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", budget)
 				setBudget = true
 			}
 			if v := tc.Get("include_thoughts"); v.Exists() {
 				out, _ = sjson.Set(out, "generationConfig.thinkingConfig.include_thoughts", v.Bool())
 			} else if setBudget {
-				if normalized != 0 {
+				if budget != 0 {
 					out, _ = sjson.Set(out, "generationConfig.thinkingConfig.include_thoughts", true)
 				}
 			}
 		}
 	}
 
-	// For gemini-3-pro-preview, always send default thinkingConfig when none specified.
-	// This matches the official Gemini CLI behavior which always sends:
-	// { thinkingBudget: -1, includeThoughts: true }
-	// See: ai-gemini-cli/packages/core/src/config/defaultModelConfigs.ts
-	if !gjson.Get(out, "generationConfig.thinkingConfig").Exists() && modelName == "gemini-3-pro-preview" {
-		out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", -1)
-		out, _ = sjson.Set(out, "generationConfig.thinkingConfig.include_thoughts", true)
-		// log.Debugf("Applied default thinkingConfig for gemini-3-pro-preview (matches Gemini CLI): thinkingBudget=-1, include_thoughts=true")
-	}
-
 	result := []byte(out)
 	result = common.AttachDefaultSafetySettings(result, "safetySettings")
 	return result

internal/util/gemini_thinking.go

@@ -207,6 +207,47 @@ func GeminiThinkingFromMetadata(metadata map[string]any) (*int, *bool, bool) {
 	return budgetPtr, includePtr, matched
 }
 
+// modelsWithDefaultThinking lists models that should have thinking enabled by default
+// when no explicit thinkingConfig is provided.
+var modelsWithDefaultThinking = map[string]bool{
+	"gemini-3-pro-preview": true,
+}
+
+// ModelHasDefaultThinking returns true if the model should have thinking enabled by default.
+func ModelHasDefaultThinking(model string) bool {
+	return modelsWithDefaultThinking[model]
+}
+
+// ApplyDefaultThinkingIfNeeded injects default thinkingConfig for models that require it.
+// For standard Gemini API format (generationConfig.thinkingConfig path).
+// Returns the modified body if thinkingConfig was added, otherwise returns the original.
+func ApplyDefaultThinkingIfNeeded(model string, body []byte) []byte {
+	if !ModelHasDefaultThinking(model) {
+		return body
+	}
+	if gjson.GetBytes(body, "generationConfig.thinkingConfig").Exists() {
+		return body
+	}
+	updated, _ := sjson.SetBytes(body, "generationConfig.thinkingConfig.thinkingBudget", -1)
+	updated, _ = sjson.SetBytes(updated, "generationConfig.thinkingConfig.include_thoughts", true)
+	return updated
+}
+
+// ApplyDefaultThinkingIfNeededCLI injects default thinkingConfig for models that require it.
+// For Gemini CLI API format (request.generationConfig.thinkingConfig path).
+// Returns the modified body if thinkingConfig was added, otherwise returns the original.
+func ApplyDefaultThinkingIfNeededCLI(model string, body []byte) []byte {
+	if !ModelHasDefaultThinking(model) {
+		return body
+	}
+	if gjson.GetBytes(body, "request.generationConfig.thinkingConfig").Exists() {
+		return body
+	}
+	updated, _ := sjson.SetBytes(body, "request.generationConfig.thinkingConfig.thinkingBudget", -1)
+	updated, _ = sjson.SetBytes(updated, "request.generationConfig.thinkingConfig.include_thoughts", true)
+	return updated
+}
+
 // StripThinkingConfigIfUnsupported removes thinkingConfig from the request body
 // when the target model does not advertise Thinking capability. It cleans both
 // standard Gemini and Gemini CLI JSON envelopes. This acts as a final safety net
@@ -223,6 +264,32 @@ func StripThinkingConfigIfUnsupported(model string, body []byte) []byte {
 	return updated
 }
 
+// NormalizeGeminiThinkingBudget normalizes the thinkingBudget value in a standard Gemini
+// request body (generationConfig.thinkingConfig.thinkingBudget path).
+func NormalizeGeminiThinkingBudget(model string, body []byte) []byte {
+	const budgetPath = "generationConfig.thinkingConfig.thinkingBudget"
+	budget := gjson.GetBytes(body, budgetPath)
+	if !budget.Exists() {
+		return body
+	}
+	normalized := NormalizeThinkingBudget(model, int(budget.Int()))
+	updated, _ := sjson.SetBytes(body, budgetPath, normalized)
+	return updated
+}
+
+// NormalizeGeminiCLIThinkingBudget normalizes the thinkingBudget value in a Gemini CLI
+// request body (request.generationConfig.thinkingConfig.thinkingBudget path).
+func NormalizeGeminiCLIThinkingBudget(model string, body []byte) []byte {
+	const budgetPath = "request.generationConfig.thinkingConfig.thinkingBudget"
+	budget := gjson.GetBytes(body, budgetPath)
+	if !budget.Exists() {
+		return body
+	}
+	normalized := NormalizeThinkingBudget(model, int(budget.Int()))
+	updated, _ := sjson.SetBytes(body, budgetPath, normalized)
+	return updated
+}
+
 // ConvertThinkingLevelToBudget checks for "generationConfig.thinkingConfig.thinkingLevel"
 // and converts it to "thinkingBudget".
 // "high" -> 32768