fix(bedrock): map reasoning_effort to reasoning.effort for OpenAI GPT-5.x on Converse

litellm/llms/bedrock/chat/converse_transformation.py

@@ -418,12 +418,16 @@ def _handle_reasoning_effort_parameter(self, model: str, reasoning_effort: str,
         Handle the reasoning_effort parameter based on the model type.
 
         - GPT-OSS models: passed through unchanged via additionalModelRequestFields.
+        - OpenAI GPT-5.x models: mapped to ``reasoning.effort`` via additionalModelRequestFields.
         - Nova 2 models: transformed to reasoningConfig.
         - Anthropic models: mapped to ``thinking`` (and ``output_config.effort`` on
           adaptive Claude 4.6 / 4.7).
         """
         if "gpt-oss" in model:
             optional_params["reasoning_effort"] = reasoning_effort
+        elif "openai.gpt-5" in model:
+            reasoning: Final[BedrockConverseGptReasoningEffortBlock] = {"effort": reasoning_effort}
+            optional_params["reasoning"] = reasoning
         elif self._is_nova_2_model(model):
             reasoning_config: Final = self._transform_reasoning_effort_to_reasoning_config(reasoning_effort)
             optional_params.update(reasoning_config)
@@ -555,7 +559,7 @@ def get_supported_openai_params(self, model: str) -> list[str]:
             # only anthropic and mistral support tool choice config. otherwise (E.g. cohere) will fail the call - https://docs.aws.amazon.com/bedrock/latest/APIReference/API_runtime_ToolChoice.html
             supported_params.append("tool_choice")
 
-        if "gpt-oss" in model:
+        if "gpt-oss" in model or "openai.gpt-5" in model or "openai.gpt-5" in base_model:
             supported_params.append("reasoning_effort")
         elif self._is_nova_2_model(model):
             # Nova 2 models support reasoning_effort (transformed to reasoningConfig)
@@ -903,7 +907,7 @@ def map_openai_params(
                 optional_params["_parallel_tool_use_config"] = {
                     "tool_choice": {"type": "auto", "disable_parallel_tool_use": not value}
                 }
-            if param == "thinking":
+            if param == "thinking" and "openai.gpt-5" not in model:
                 if (
                     isinstance(value, dict)
                     and value.get("type") == "adaptive"

litellm/model_prices_and_context_window_backup.json

@@ -49473,6 +49473,7 @@
         ],
         "supports_function_calling": true,
         "supports_tool_choice": true,
+        "supports_reasoning": true,
         "supports_vision": true
     },
     "global.openai.gpt-5.6-sol": {
@@ -49498,6 +49499,7 @@
         ],
         "supports_function_calling": true,
         "supports_tool_choice": true,
+        "supports_reasoning": true,
         "supports_vision": true
     },
     "us.openai.gpt-5.6-terra": {
@@ -49523,6 +49525,7 @@
         ],
         "supports_function_calling": true,
         "supports_tool_choice": true,
+        "supports_reasoning": true,
         "supports_vision": true
     },
     "global.openai.gpt-5.6-terra": {
@@ -49548,6 +49551,7 @@
         ],
         "supports_function_calling": true,
         "supports_tool_choice": true,
+        "supports_reasoning": true,
         "supports_vision": true
     },
     "us.openai.gpt-5.6-luna": {
@@ -49573,6 +49577,7 @@
         ],
         "supports_function_calling": true,
         "supports_tool_choice": true,
+        "supports_reasoning": true,
         "supports_vision": true
     },
     "global.openai.gpt-5.6-luna": {
@@ -49598,6 +49603,7 @@
         ],
         "supports_function_calling": true,
         "supports_tool_choice": true,
+        "supports_reasoning": true,
         "supports_vision": true
     },
     "bedrock_mantle/openai.gpt-5.5": {

litellm/types/llms/bedrock.py

@@ -3,7 +3,7 @@
 from enum import Enum
 from typing import TYPE_CHECKING, Any, Final, Literal
 
-from typing_extensions import Required, TypedDict, override
+from typing_extensions import ReadOnly, Required, TypedDict, override
 
 from .openai import ChatCompletionToolCallChunk
 
@@ -97,6 +97,10 @@ class BedrockConverseReasoningContentBlockDelta(TypedDict, total=False):
     text: str
 
 
+class BedrockConverseGptReasoningEffortBlock(TypedDict):
+    effort: ReadOnly[str]
+
+
 class GuardrailConverseTextBlock(TypedDict, total=False):
     text: str
 

model_prices_and_context_window.json

@@ -49473,6 +49473,7 @@
         ],
         "supports_function_calling": true,
         "supports_tool_choice": true,
+        "supports_reasoning": true,
         "supports_vision": true
     },
     "global.openai.gpt-5.6-sol": {
@@ -49498,6 +49499,7 @@
         ],
         "supports_function_calling": true,
         "supports_tool_choice": true,
+        "supports_reasoning": true,
         "supports_vision": true
     },
     "us.openai.gpt-5.6-terra": {
@@ -49523,6 +49525,7 @@
         ],
         "supports_function_calling": true,
         "supports_tool_choice": true,
+        "supports_reasoning": true,
         "supports_vision": true
     },
     "global.openai.gpt-5.6-terra": {
@@ -49548,6 +49551,7 @@
         ],
         "supports_function_calling": true,
         "supports_tool_choice": true,
+        "supports_reasoning": true,
         "supports_vision": true
     },
     "us.openai.gpt-5.6-luna": {
@@ -49573,6 +49577,7 @@
         ],
         "supports_function_calling": true,
         "supports_tool_choice": true,
+        "supports_reasoning": true,
         "supports_vision": true
     },
     "global.openai.gpt-5.6-luna": {
@@ -49598,6 +49603,7 @@
         ],
         "supports_function_calling": true,
         "supports_tool_choice": true,
+        "supports_reasoning": true,
         "supports_vision": true
     },
     "bedrock_mantle/openai.gpt-5.5": {

tests/test_litellm/llms/bedrock/chat/test_converse_transformation.py

@@ -284,6 +284,71 @@ def test_reasoning_with_forced_tool_choice_switches_to_auto():
     assert optional_params["tool_choice"] == {"auto": {}}
 
 
+@pytest.mark.parametrize(
+    "model",
+    [
+        "us.openai.gpt-5.6-sol",
+        "global.openai.gpt-5.6-terra",
+        "bedrock/converse/us.openai.gpt-5.6-luna",
+    ],
+)
+def test_reasoning_effort_maps_to_reasoning_effort_for_openai_gpt5_converse(model, local_model_cost_map):
+    """OpenAI GPT-5.x on Bedrock Converse routes reasoning_effort to
+    ``additionalModelRequestFields.reasoning.effort`` rather than Anthropic ``thinking``."""
+    config = AmazonConverseConfig()
+
+    assert "reasoning_effort" in config.get_supported_openai_params(model=model)
+
+    optional_params = config.map_openai_params(
+        non_default_params={"reasoning_effort": "high"},
+        optional_params={},
+        model=model,
+        drop_params=False,
+    )
+
+    assert optional_params["reasoning"] == {"effort": "high"}
+    assert "thinking" not in optional_params
+    assert "reasoning_effort" not in optional_params
+
+    _, additional_request_params, _, _ = config._prepare_request_params(optional_params, model)
+    assert additional_request_params["reasoning"] == {"effort": "high"}
+    assert "thinking" not in additional_request_params
+
+
+@pytest.mark.parametrize(
+    "model",
+    [
+        "us.openai.gpt-5.6-sol",
+        "bedrock/converse/global.openai.gpt-5.6-luna",
+    ],
+)
+def test_openai_gpt5_converse_never_forwards_thinking(model, local_model_cost_map):
+    """GPT-5.x on Converse must never send Anthropic ``thinking``/``output_config`` (Bedrock rejects them).
+
+    Regression: ``thinking`` is not advertised as supported, and even when supplied alongside
+    ``reasoning_effort`` in either order it never survives into the request."""
+    config = AmazonConverseConfig()
+
+    supported = config.get_supported_openai_params(model=model)
+    assert "thinking" not in supported
+    assert "output_config" not in supported
+
+    thinking_block = {"type": "enabled", "budget_tokens": 2048}
+    for non_default_params in (
+        {"reasoning_effort": "high", "thinking": thinking_block},
+        {"thinking": thinking_block, "reasoning_effort": "high"},
+    ):
+        optional_params = config.map_openai_params(
+            non_default_params=dict(non_default_params),
+            optional_params={},
+            model=model,
+            drop_params=False,
+        )
+        _, additional_request_params, _, _ = config._prepare_request_params(optional_params, model)
+        assert additional_request_params["reasoning"] == {"effort": "high"}
+        assert "thinking" not in additional_request_params
+
+
 @pytest.mark.parametrize(
     "model",
     [

tests/test_litellm/llms/bedrock/test_cross_region_inference_profile_mapping.py

@@ -293,15 +293,16 @@ def test_bedrock_gpt_5_6_advertises_only_converse_supported_features(
 
 
 @pytest.mark.parametrize("profile", GPT_5_6_PROFILES, ids=lambda p: p.model_id)
-def test_bedrock_gpt_5_6_offers_tools_but_not_reasoning(profile, local_model_cost_map):
-    """Converse rejects the Anthropic-shaped thinking block LiteLLM emits for
-    reasoning_effort, so neither reasoning param may be offered yet, while the tool
-    params these models do accept must be."""
+def test_bedrock_gpt_5_6_offers_tools_and_reasoning_effort_but_not_thinking(profile, local_model_cost_map):
+    """GPT-5.x on Converse maps reasoning_effort to reasoning.effort, so reasoning_effort
+    is offered while the Anthropic-only thinking/output_config are not, alongside the tool
+    params these models accept."""
     supported = AmazonConverseConfig().get_supported_openai_params(
         model=f"bedrock/{profile.model_id}"
     )
 
     assert "tools" in supported
     assert "tool_choice" in supported
-    assert "reasoning_effort" not in supported
+    assert "reasoning_effort" in supported
     assert "thinking" not in supported
+    assert "output_config" not in supported