diff --git a/src/renderer/src/config/models.ts b/src/renderer/src/config/models.ts
index 0b0afc398b..3ae1814f59 100644
--- a/src/renderer/src/config/models.ts
+++ b/src/renderer/src/config/models.ts
@@ -2631,7 +2631,8 @@ export function groupQwenModels(models: Model[]): Record<string, Model[]> {
 
 export const THINKING_TOKEN_MAP: Record<string, { min: number; max: number }> = {
   // Gemini models
-  'gemini-.*$': { min: 0, max: 24576 },
+  'gemini-.*-flash.*$': { min: 0, max: 24576 },
+  'gemini-.*-pro.*$': { min: 128, max: 32768 },
 
   // Qwen models
   'qwen-plus-.*$': { min: 0, max: 38912 },
diff --git a/src/renderer/src/providers/AiProvider/GeminiProvider.ts b/src/renderer/src/providers/AiProvider/GeminiProvider.ts
index 0186beca14..97c90363d0 100644
--- a/src/renderer/src/providers/AiProvider/GeminiProvider.ts
+++ b/src/renderer/src/providers/AiProvider/GeminiProvider.ts
@@ -286,12 +286,14 @@ export default class GeminiProvider extends BaseProvider {
   private getBudgetToken(assistant: Assistant, model: Model) {
     if (isGeminiReasoningModel(model)) {
       const reasoningEffort = assistant?.settings?.reasoning_effort
+      const GEMINI_FLASH_MODEL_REGEX = new RegExp('gemini-.*-flash.*$')
 
       // 如果thinking_budget是undefined，不思考
       if (reasoningEffort === undefined) {
         return {
           thinkingConfig: {
-            includeThoughts: false
+            includeThoughts: false,
+            ...(GEMINI_FLASH_MODEL_REGEX.test(model.id) ? { thinkingBudget: 0 } : {})
           } as ThinkingConfig
         }
       }