fix(core): skip enable_thinking=false for thinking-only models

packages/core/src/core/contentGenerator.ts

@@ -137,6 +137,9 @@ export type ContentGeneratorConfig = {
   customHeaders?: Record<string, string>;
   // Extra body parameters to be merged into the request body
   extra_body?: Record<string, unknown>;
+  // When true, the model requires thinking to be enabled and rejects
+  // enable_thinking=false with a 400 error (e.g. qwen3.8-max-preview).
+  thinkingMandatory?: boolean;
   // Supported input modalities. Unsupported media types are replaced with text
   // placeholders. Leave undefined to use automatic detection from model name.
   modalities?: InputModalities;

packages/core/src/core/openaiContentGenerator/pipeline.test.ts

@@ -952,6 +952,86 @@ describe('ContentGenerationPipeline', () => {
       expect(apiCall.enable_thinking).toBe(false);
     });
 
+    it('skips enable_thinking:false for thinking-only models (#7332)', async () => {
+      // qwen3.8-max-preview rejects enable_thinking=false with a 400 error.
+      // The preset signals this via thinkingMandatory=true.
+      mockContentGeneratorConfig = {
+        ...mockContentGeneratorConfig,
+        baseUrl: 'https://dashscope.aliyuncs.com/compatible-mode/v1',
+        model: 'qwen3.8-max-preview',
+        extra_body: { enable_thinking: true },
+        thinkingMandatory: true,
+      } as ContentGeneratorConfig;
+      mockConfig = {
+        ...mockConfig,
+        contentGeneratorConfig: mockContentGeneratorConfig,
+      };
+      pipeline = new ContentGenerationPipeline(mockConfig);
+
+      const request: GenerateContentParameters = {
+        model: 'qwen3.8-max-preview',
+        contents: [{ parts: [{ text: 'Summarize' }], role: 'user' }],
+        config: { thinkingConfig: { includeThoughts: false } },
+      };
+
+      (mockConverter.convertGeminiRequestToOpenAI as Mock).mockReturnValue([
+        { role: 'user', content: 'Summarize' },
+      ]);
+      (mockConverter.convertOpenAIResponseToGemini as Mock).mockReturnValue(
+        new GenerateContentResponse(),
+      );
+      (mockClient.chat.completions.create as Mock).mockResolvedValue({
+        id: 'r',
+        choices: [{ message: { content: 'ok' }, finish_reason: 'stop' }],
+      } as OpenAI.Chat.ChatCompletion);
+
+      await pipeline.execute(request, 'forked_query');
+
+      const apiCall = (mockClient.chat.completions.create as Mock).mock
+        .calls[0][0];
+      expect(apiCall.enable_thinking).not.toBe(false);
+    });
+
+    it('emits enable_thinking:false for hybrid models with extra_body.enable_thinking (#7332)', async () => {
+      // Hybrid models (e.g. qwen3.7-max) have extra_body.enable_thinking=true
+      // from their preset but do NOT have thinkingMandatory. The pipeline must
+      // still emit enable_thinking=false when reasoning is disabled.
+      mockContentGeneratorConfig = {
+        ...mockContentGeneratorConfig,
+        baseUrl: 'https://dashscope.aliyuncs.com/compatible-mode/v1',
+        model: 'qwen3.7-max',
+        extra_body: { enable_thinking: true },
+      } as ContentGeneratorConfig;
+      mockConfig = {
+        ...mockConfig,
+        contentGeneratorConfig: mockContentGeneratorConfig,
+      };
+      pipeline = new ContentGenerationPipeline(mockConfig);
+
+      const request: GenerateContentParameters = {
+        model: 'qwen3.7-max',
+        contents: [{ parts: [{ text: 'Summarize' }], role: 'user' }],
+        config: { thinkingConfig: { includeThoughts: false } },
+      };
+
+      (mockConverter.convertGeminiRequestToOpenAI as Mock).mockReturnValue([
+        { role: 'user', content: 'Summarize' },
+      ]);
+      (mockConverter.convertOpenAIResponseToGemini as Mock).mockReturnValue(
+        new GenerateContentResponse(),
+      );
+      (mockClient.chat.completions.create as Mock).mockResolvedValue({
+        id: 'r',
+        choices: [{ message: { content: 'ok' }, finish_reason: 'stop' }],
+      } as OpenAI.Chat.ChatCompletion);
+
+      await pipeline.execute(request, 'forked_query');
+
+      const apiCall = (mockClient.chat.completions.create as Mock).mock
+        .calls[0][0];
+      expect(apiCall.enable_thinking).toBe(false);
+    });
+
     it('emits enable_thinking:false on DashScope hostname when reasoning is configured to false', async () => {
       // Config-level opt-out (`reasoning: false`) should also disable
       // qwen3 thinking, mirroring the DeepSeek pair above.

packages/core/src/core/openaiContentGenerator/pipeline.ts

@@ -847,7 +847,12 @@ export class ContentGenerationPipeline {
             this.contentGeneratorConfig,
           )
         ) {
-          typed['enable_thinking'] = false;
+          // Skip disabling thinking for thinking-only models (e.g.
+          // qwen3.8-max-preview) that reject enable_thinking=false with a
+          // 400 error (#7332). The preset signals this via thinkingMandatory.
+          if (!this.contentGeneratorConfig.thinkingMandatory) {
+            typed['enable_thinking'] = false;
+          }
         } else {
           // Non-DashScope OpenAI-compatible servers (vLLM, SGLang, ...) render
           // the model's chat template server-side and read the thinking switch

packages/core/src/models/constants.ts

@@ -30,6 +30,7 @@ export const MODEL_GENERATION_CONFIG_FIELDS = [
   'contextWindowSize',
   'customHeaders',
   'extra_body',
+  'thinkingMandatory',
   'modalities',
   'splitToolMedia',
   'toolResultContentFormat',

packages/core/src/models/types.ts

@@ -39,6 +39,7 @@ export type ModelGenerationConfig = Pick<
   | 'reasoning'
   | 'customHeaders'
   | 'extra_body'
+  | 'thinkingMandatory'
   | 'contextWindowSize'
   | 'modalities'
   | 'splitToolMedia'

packages/core/src/providers/presets/alibaba-token-plan.ts

@@ -32,6 +32,7 @@ const TOKEN_PLAN_MODELS: ModelSpec[] = [
   {
     id: 'qwen3.8-max-preview',
     contextWindowSize: 1000000,
+    thinkingMandatory: true,
     enableThinking: true,
     modalities: { image: true, video: true },
   },

packages/core/src/providers/provider-config.ts

@@ -59,14 +59,21 @@ export function resolveOwnsModel(
 }
 
 function buildGenerationConfig(
-  spec: Pick<ModelSpec, 'enableThinking' | 'contextWindowSize' | 'modalities'>,
+  spec: Pick<
+    ModelSpec,
+    'enableThinking' | 'thinkingMandatory' | 'contextWindowSize' | 'modalities'
+  >,
 ): ProviderModelConfig['generationConfig'] | undefined {
   const parts: ProviderModelConfig['generationConfig'] = {};
   let hasAny = false;
   if (spec.enableThinking) {
     parts.extra_body = { enable_thinking: true };
     hasAny = true;
   }
+  if (spec.thinkingMandatory) {
+    parts.thinkingMandatory = true;
+    hasAny = true;
+  }
   if (spec.contextWindowSize) {
     parts.contextWindowSize = spec.contextWindowSize;
     hasAny = true;

packages/core/src/providers/types.ts

@@ -20,6 +20,7 @@ export interface ModelSpec {
   id: string;
   contextWindowSize?: number;
   enableThinking?: boolean;
+  thinkingMandatory?: boolean;
   modalities?: InputModalities;
   description?: string;
 }