diff --git a/app/services/llm_service.py b/app/services/llm_service.py
index e955f87..7ba8668 100644
--- a/app/services/llm_service.py
+++ b/app/services/llm_service.py
@@ -72,7 +72,20 @@ def _complete(self, model: Optional[str], messages: List[dict], temperature: flo
         resolved = self._resolve_model(model)
         self._validate_model_environment(resolved)
         logger.debug("Requesting completion from model: %s", resolved)
-        return completion(model=resolved, messages=messages, temperature=temperature)
+        try:
+            return completion(model=resolved, messages=messages, temperature=temperature)
+        except litellm.BadRequestError as e:
+            # Some models (e.g. OpenAI's o1 / gpt-5 series) only accept the default
+            # temperature and reject any explicit value. Retry once letting the
+            # model use its own default rather than failing the whole request.
+            if "temperature" in str(e).lower():
+                logger.warning(
+                    "Model '%s' rejected temperature=%s; retrying with the model default.",
+                    resolved,
+                    temperature,
+                )
+                return completion(model=resolved, messages=messages)
+            raise
 
     def get_artist_recommendations(self, prompt: str, artists: List[Artist], model: str = None):
         """First step: Get relevant artists based on the prompt"""
diff --git a/tests/test_llm_service.py b/tests/test_llm_service.py
index 3c7da3c..c717d91 100644
--- a/tests/test_llm_service.py
+++ b/tests/test_llm_service.py
@@ -173,3 +173,45 @@ def test_model_defaults_to_configured_value(mock_completion, sample_artists, mon
     _, kwargs = mock_completion.call_args
     assert kwargs["model"] == "gpt-4"
     config.get_settings.cache_clear()
+
+
+def test_complete_retries_without_temperature_when_unsupported(mock_completion, sample_artists):
+    """Models that reject a custom temperature are retried with the model default."""
+    import litellm  # pylint: disable=import-outside-toplevel
+
+    error = litellm.BadRequestError(
+        "OpenAIException - Unsupported value: 'temperature' does not support 0.7 with this model.",
+        model="gpt-5.6-sol",
+        llm_provider="openai",
+    )
+    good_response = Mock()
+    good_response.choices = [Mock(message=Mock(content='{"artists": ["Artist1"]}'))]
+
+    calls = []
+
+    def side_effect(*args, **kwargs):  # pylint: disable=unused-argument
+        calls.append(kwargs)
+        if "temperature" in kwargs:
+            raise error
+        return good_response
+
+    mock_completion.side_effect = side_effect
+
+    service = LLMService()
+    result = service.get_artist_recommendations("Test prompt", sample_artists, model="gpt-5.6-sol")
+
+    assert result == ["Artist1"]
+    assert len(calls) == 2  # first attempt (with temperature) failed, retry succeeded
+    assert "temperature" not in calls[1]
+
+
+def test_complete_reraises_unrelated_bad_request(mock_completion, sample_artists):
+    """A BadRequestError unrelated to temperature is not swallowed."""
+    import litellm  # pylint: disable=import-outside-toplevel
+
+    error = litellm.BadRequestError("context_length_exceeded", model="gpt-4", llm_provider="openai")
+    mock_completion.side_effect = error
+
+    service = LLMService()
+    with pytest.raises(litellm.BadRequestError, match="context_length_exceeded"):
+        service.get_artist_recommendations("Test prompt", sample_artists, model="gpt-4")