diff --git a/packages/bruno-electron/src/ipc/ai/autocomplete.js b/packages/bruno-electron/src/ipc/ai/autocomplete.js
index f214b37f2..a665b6bb5 100644
Binary files a/packages/bruno-electron/src/ipc/ai/autocomplete.js and b/packages/bruno-electron/src/ipc/ai/autocomplete.js differ
diff --git a/packages/bruno-electron/src/ipc/ai/chat.js b/packages/bruno-electron/src/ipc/ai/chat.js
index 67278b815..5ca7c98a5 100644
--- a/packages/bruno-electron/src/ipc/ai/chat.js
+++ b/packages/bruno-electron/src/ipc/ai/chat.js
@@ -245,6 +245,7 @@ const registerChatIpc = ({ mainWindow, resolveModel, pickDefaultModelId, isAiEna
         abortSignal: controller.signal
       });
 
+      let streamError = null;
       for await (const part of result.fullStream) {
         if (controller.signal.aborted) break;
         switch (part.type) {
@@ -271,11 +272,17 @@ const registerChatIpc = ({ mainWindow, resolveModel, pickDefaultModelId, isAiEna
             send('main:ai-chat-tool-done', { requestId, toolName: part.toolName });
             break;
           }
+          case 'error': {
+            streamError = part.error;
+            break;
+          }
           default:
             break;
         }
       }
 
+      if (streamError) throw streamError;
+
       activeStreams.delete(requestId);
 
       if (controller.signal.aborted) {
diff --git a/packages/bruno-electron/src/ipc/ai/index.js b/packages/bruno-electron/src/ipc/ai/index.js
index a2e46305f..e8ca4998c 100644
--- a/packages/bruno-electron/src/ipc/ai/index.js
+++ b/packages/bruno-electron/src/ipc/ai/index.js
@@ -12,6 +12,7 @@ const {
   clearSdkCache,
   isKnownProviderId,
   isBuiltInModelId,
+  isReasoningModel,
   validateApiKeyForProvider,
   providerLabel
 } = require('./providers');
@@ -162,7 +163,7 @@ const registerAiIpc = (mainWindow) => {
         system,
         prompt,
         maxOutputTokens: maxTokens ?? 1024,
-        temperature: temperature ?? 0.3
+        ...(isReasoningModel(modelId) ? {} : { temperature: temperature ?? 0.3 })
       });
       return { text };
     } catch (err) {
@@ -344,7 +345,7 @@ const registerAiIpc = (mainWindow) => {
         model,
         system,
         maxOutputTokens: maxTokens ?? 2048,
-        temperature: temperature ?? 0.7,
+        ...(isReasoningModel(modelId) ? {} : { temperature: temperature ?? 0.7 }),
         abortSignal: controller.signal
       };
       if (messages) {
diff --git a/packages/bruno-electron/src/ipc/ai/providers.js b/packages/bruno-electron/src/ipc/ai/providers.js
index 44dc5a8d4..ea8274c5f 100644
--- a/packages/bruno-electron/src/ipc/ai/providers.js
+++ b/packages/bruno-electron/src/ipc/ai/providers.js
@@ -40,16 +40,27 @@ const PROVIDERS = {
  * Static model catalog for built-in providers. User-defined custom models for
  * OpenAI-compatible endpoints are layered on top at lookup time.
  */
+// `reasoning: true` marks models whose SDK path rejects temperature/stopSequences
+// (OpenAI Responses API reasoning models) or accepts them only when thinking is
+// off (Anthropic Claude 4+). Callers doing latency-critical work like
+// autocomplete drop those params for reasoning models to silence warnings.
 const MODEL_DEFINITIONS = {
   // OpenAI
   'gpt-4o-mini': { provider: 'openai', modelId: 'gpt-4o-mini', label: 'GPT-4o Mini' },
   'gpt-4o': { provider: 'openai', modelId: 'gpt-4o', label: 'GPT-4o' },
-  'gpt-5': { provider: 'openai', modelId: 'gpt-5', label: 'GPT-5' },
-  'gpt-5-mini': { provider: 'openai', modelId: 'gpt-5-mini', label: 'GPT-5 Mini' },
+  'gpt-5': { provider: 'openai', modelId: 'gpt-5', label: 'GPT-5', reasoning: true },
+  'gpt-5-mini': { provider: 'openai', modelId: 'gpt-5-mini', label: 'GPT-5 Mini', reasoning: true },
   // Anthropic
-  'claude-opus-4-7': { provider: 'anthropic', modelId: 'claude-opus-4-7', label: 'Claude Opus 4.7' },
-  'claude-sonnet-4-6': { provider: 'anthropic', modelId: 'claude-sonnet-4-6', label: 'Claude Sonnet 4.6' },
-  'claude-haiku-4-5': { provider: 'anthropic', modelId: 'claude-haiku-4-5-20251001', label: 'Claude Haiku 4.5' }
+  'claude-opus-4-7': { provider: 'anthropic', modelId: 'claude-opus-4-7', label: 'Claude Opus 4.7', reasoning: true },
+  'claude-sonnet-4-6': { provider: 'anthropic', modelId: 'claude-sonnet-4-6', label: 'Claude Sonnet 4.6', reasoning: true },
+  'claude-haiku-4-5': { provider: 'anthropic', modelId: 'claude-haiku-4-5-20251001', label: 'Claude Haiku 4.5', reasoning: true }
+};
+
+const isReasoningModel = (modelId) => Boolean(MODEL_DEFINITIONS[modelId]?.reasoning);
+
+const isOpenAiReasoningModel = (modelId) => {
+  const def = MODEL_DEFINITIONS[modelId];
+  return Boolean(def?.reasoning && def?.provider === 'openai');
 };
 
 // Cache SDK instances. Built-in keyed by `${providerId}:${apiKey}`; compat
@@ -208,7 +219,9 @@ const getModel = (modelId, { aiPreferences, getApiKey }) => {
     throw new Error(`${providerLabel(def.providerId, aiPreferences)} is missing a Base URL. Set one in Preferences > AI.`);
   }
 
-  return getSdk({ providerId: def.providerId, apiKey, baseURL: def.baseURL })(def.sdkModelId);
+  const sdk = getSdk({ providerId: def.providerId, apiKey, baseURL: def.baseURL });
+  if (isOpenAiCompatibleProviderId(def.providerId)) return sdk.chat(def.sdkModelId);
+  return sdk(def.sdkModelId);
 };
 
 /**
@@ -273,6 +286,8 @@ module.exports = {
   getCompatEndpoint,
   isKnownProviderId,
   isBuiltInModelId,
+  isReasoningModel,
+  isOpenAiReasoningModel,
   validateApiKeyForProvider,
   providerLabel
 };