Reserve reasoning for responses via chat completion

litellm/llms/openai/chat/gpt_5_transformation.py

@@ -188,11 +188,9 @@ class OpenAIGPT5Config(OpenAIGPTConfig):
         ) or optional_params.get("reasoning_effort")
         effective_effort = _get_effort_level(raw_reasoning_effort)
 
-        # Normalize to string for Chat Completions API when dict has only "effort".
-        # Preserve full dict (e.g. {"effort": "high", "summary": "detailed"}) for Responses API.
-        if isinstance(raw_reasoning_effort, dict) and set(
-            raw_reasoning_effort.keys()
-        ) <= {"effort"}:
+        # Normalize dict reasoning_effort to string for Chat Completions API.
+        # Example: {"effort": "high", "summary": "detailed"} -> "high"
+        if isinstance(raw_reasoning_effort, dict) and "effort" in raw_reasoning_effort:
             normalized = _normalize_reasoning_effort_for_chat_completion(
                 raw_reasoning_effort
             )

litellm/main.py

@@ -955,7 +955,7 @@ def responses_api_bridge_check(
             model_info["mode"] = "responses"
             model = model.replace("responses/", "")
 
-        # OpenAI gpt-5.4 chat-completions calls with both tools + reasoning_effort
+        # OpenAI gpt-5.4+ chat-completions calls with both tools + reasoning_effort
         # must be bridged to Responses API.
         if (
             custom_llm_provider == "openai"
@@ -1617,6 +1617,10 @@ def completion(  # type: ignore # noqa: PLR0915
         if model_info.get("mode") == "responses":
             from litellm.completion_extras import responses_api_bridge
 
+            if isinstance(reasoning_effort, dict) and "summary" in reasoning_effort:
+                optional_params = dict(optional_params)
+                optional_params["reasoning_effort"] = reasoning_effort
+
             return responses_api_bridge.completion(
                 model=model,
                 messages=messages,

tests/test_litellm/llms/openai/test_gpt5_transformation.py

@@ -324,19 +324,15 @@ def test_gpt5_4_pro_allows_reasoning_effort_xhigh(config: OpenAIConfig):
     assert params["reasoning_effort"] == "xhigh"
 
 
-def test_gpt5_preserves_reasoning_effort_dict_with_summary(config: OpenAIConfig):
-    """Dict with summary/generate_summary is preserved for Responses API.
-
-    Config/deployments may pass Responses API format: {'effort': 'high', 'summary': 'detailed'}.
-    We preserve the full dict so it reaches the Responses API transformation.
-    """
+def test_gpt5_normalizes_reasoning_effort_dict_with_summary(config: OpenAIConfig):
+    """Dict with summary/generate_summary is normalized for chat completions."""
     params = config.map_openai_params(
         non_default_params={"reasoning_effort": {"effort": "high", "summary": "detailed"}},
         optional_params={},
         model="gpt-5.4",
         drop_params=False,
     )
-    assert params["reasoning_effort"] == {"effort": "high", "summary": "detailed"}
+    assert params["reasoning_effort"] == "high"
 
 
 def test_gpt5_xhigh_dict_triggers_validation(config: OpenAIConfig):
@@ -362,7 +358,7 @@ def test_gpt5_xhigh_dict_accepted_for_supported_model(config: OpenAIConfig):
         model="gpt-5.4",
         drop_params=False,
     )
-    assert params["reasoning_effort"] == {"effort": "xhigh", "summary": "detailed"}
+    assert params["reasoning_effort"] == "xhigh"
 
 
 def test_gpt5_none_dict_with_tools_no_tool_drop(config: OpenAIConfig):
@@ -378,7 +374,7 @@ def test_gpt5_none_dict_with_tools_no_tool_drop(config: OpenAIConfig):
         model="gpt-5.4",
         drop_params=False,
     )
-    assert params["reasoning_effort"] == {"effort": "none", "summary": "detailed"}
+    assert params["reasoning_effort"] == "none"
     assert params["tools"] == tools
 
 
@@ -398,20 +394,20 @@ def test_gpt5_none_dict_with_sampling_params_allowed(config: OpenAIConfig):
         model="gpt-5.1",
         drop_params=False,
     )
-    assert params["reasoning_effort"] == {"effort": "none", "summary": "detailed"}
+    assert params["reasoning_effort"] == "none"
     assert params["logprobs"] is True
     assert params["top_p"] == 0.9
 
 
-def test_gpt5_preserves_reasoning_effort_dict_with_summary_from_optional_params(config: OpenAIConfig):
-    """reasoning_effort dict with summary in optional_params is preserved."""
+def test_gpt5_normalizes_reasoning_effort_dict_with_summary_from_optional_params(config: OpenAIConfig):
+    """reasoning_effort dict with summary in optional_params is normalized."""
     params = config.map_openai_params(
         non_default_params={},
         optional_params={"reasoning_effort": {"effort": "medium", "summary": "detailed"}},
         model="gpt-5.4",
         drop_params=False,
     )
-    assert params["reasoning_effort"] == {"effort": "medium", "summary": "detailed"}
+    assert params["reasoning_effort"] == "medium"
 
 
 def test_gpt5_4_pro_rejects_non_default_temperature(config: OpenAIConfig):

tests/test_litellm/test_main.py

@@ -678,6 +678,43 @@ def test_responses_api_bridge_check_gpt_5_4_tools_without_reasoning_stays_chat()
     assert model_info.get("mode") != "responses"
 
 
+@patch("litellm.completion_extras.responses_api_bridge.completion")
+def test_gpt_5_4_responses_bridge_preserves_reasoning_summary_dict(
+    mock_responses_completion,
+):
+    """When routed to Responses, preserve reasoning_effort summary dict."""
+    mock_responses_completion.return_value = MagicMock()
+
+    import litellm
+
+    litellm.completion(
+        model="gpt-5.4",
+        messages=[{"role": "user", "content": "What is the capital of France?"}],
+        tools=[
+            {
+                "type": "function",
+                "function": {
+                    "name": "get_capital",
+                    "description": "Get the capital of a country",
+                    "parameters": {
+                        "type": "object",
+                        "properties": {"country": {"type": "string"}},
+                    },
+                },
+            }
+        ],
+        reasoning_effort={"effort": "xhigh", "summary": "detailed"},
+        api_key="fake-key",
+    )
+
+    assert mock_responses_completion.called is True
+    optional_params = mock_responses_completion.call_args.kwargs["optional_params"]
+    assert optional_params["reasoning_effort"] == {
+        "effort": "xhigh",
+        "summary": "detailed",
+    }
+
+
 def test_responses_api_bridge_check_handles_exception():
     """Test that responses_api_bridge_check handles exceptions and still processes responses/ models."""
     from litellm.main import responses_api_bridge_check