diff --git a/src/utils/ai.js b/src/utils/ai.js
index cafd1de1..62048fcb 100644
--- a/src/utils/ai.js
+++ b/src/utils/ai.js
@@ -7,7 +7,7 @@
 const { OpenAI } = require('openai');
 const { DEFAULT_MODEL, FALLBACK_MODEL, getBedrockModelId, getBedrockMaxOut } = require('./models');
 const logger = require('./logger');
-const { resolveGenerationOptions } = require('./generationOptions');
+const { resolveGenerationOptions, addReasoningOptions } = require('./generationOptions');
 
 var activeProvider = process.env.AI_PROVIDER || (process.env.LITELLM_API_BASE ? 'litellm' : 'openrouter');
 
@@ -368,12 +368,6 @@ async function callVertex(messages, model, temperature, maxTokens) {
 // ============================================================
 // CALL LITELLM (OpenAI-compatible proxy)
 // ============================================================
-function addReasoningOptions(request, generation) {
-  if (generation.reasoningEffort != null) request.reasoning_effort = generation.reasoningEffort;
-  if (generation.reasoningFormat != null) request.reasoning_format = generation.reasoningFormat;
-  return request;
-}
-
 async function callLiteLLM(messages, model, temperature, maxTokens, generation) {
   if (!litellmClient) throw new Error('LiteLLM not configured. Set LITELLM_API_BASE in .env');
 
diff --git a/src/utils/generationOptions.js b/src/utils/generationOptions.js
index 21705a14..5aa0ef3e 100644
--- a/src/utils/generationOptions.js
+++ b/src/utils/generationOptions.js
@@ -8,4 +8,13 @@ function resolveGenerationOptions(options) {
   };
 }
 
-module.exports = { resolveGenerationOptions };
+function addReasoningOptions(request, generation) {
+  // These OpenAI-compatible fields are rejected by ordinary chat models such as GPT-4.1.
+  // Keep the Clinical Assistant profile, but only send it to the Groq Qwen model that supports both fields.
+  if (request.model !== 'groq-qwen3.8-27b') return request;
+  if (generation.reasoningEffort != null) request.reasoning_effort = generation.reasoningEffort;
+  if (generation.reasoningFormat != null) request.reasoning_format = generation.reasoningFormat;
+  return request;
+}
+
+module.exports = { resolveGenerationOptions, addReasoningOptions };
diff --git a/test/clinical-generation-options.test.js b/test/clinical-generation-options.test.js
index 2f85a7e5..800ca1d7 100644
--- a/test/clinical-generation-options.test.js
+++ b/test/clinical-generation-options.test.js
@@ -2,7 +2,7 @@ const test = require('node:test');
 const assert = require('node:assert/strict');
 
 const { assistantGenerationOptions } = require('../src/utils/clinicalAnswer');
-const { resolveGenerationOptions } = require('../src/utils/generationOptions');
+const { resolveGenerationOptions, addReasoningOptions } = require('../src/utils/generationOptions');
 
 test('clinical assistant profile requests low reasoning without exposing it', () => {
   assert.deepEqual(assistantGenerationOptions({ temperature: 0, maxTokens: 80 }), {
@@ -17,3 +17,11 @@ test('generation defaults preserve an explicit zero temperature', () => {
   assert.equal(resolveGenerationOptions({ temperature: 0 }).temperature, 0);
   assert.equal(resolveGenerationOptions({}).temperature, 0.3);
 });
+
+test('reasoning fields are sent only to the confirmed Groq Qwen model', () => {
+  const profile = { reasoningEffort: 'low', reasoningFormat: 'hidden' };
+  assert.deepEqual(addReasoningOptions({ model: 'openai-gpt-4.1' }, profile), { model: 'openai-gpt-4.1' });
+  assert.deepEqual(addReasoningOptions({ model: 'groq-qwen3.8-27b' }, profile), {
+    model: 'groq-qwen3.8-27b', reasoning_effort: 'low', reasoning_format: 'hidden'
+  });
+});