fix: support served_model_name for Baseten dedicated deployments

docs/my-website/docs/providers/baseten.md

@@ -82,6 +82,8 @@ for chunk in response:
 
 ## Usage with LiteLLM Proxy
 
+### Model API
+
 1. **Config**:
 ```yaml
 model_list:
@@ -91,16 +93,23 @@ model_list:
       api_key: your-baseten-api-key
 ```
 
-2. **Request**:
-```python
-import openai
-client = openai.OpenAI(
-    api_key="sk-1234",
-    base_url="http://0.0.0.0:4000"
-)
+### Dedicated Deployment
 
-response = client.chat.completions.create(
-    model="baseten-model",
-    messages=[{"role": "user", "content": "Hello!"}]
-)
+If your dedicated deployment uses a `served_model_name` in your Baseten `config.yaml`, you must supply `served_model_name` to specify the model name sent in the request body, and supply the model id under the `model` field.
+
+1. **Config**:
+```yaml
+model_list:
+  - model_name: baseten-model # external user facing
+    litellm_params:
+      model: baseten/1234abcd # model id from Baseten dashboard
+      served_model_name: baseten-hosted/zai-org/GLM-5 # model name specified in Baseten config.yaml
+      api_key: os.environ/BASETEN_API_KEY
 ```
+
+- `model: baseten/1234abcd` — the 8-digit deployment ID, used to route to `https://model-1234abcd.api.baseten.co/environments/production/sync/v1`
+- `served_model_name` — sent as the `model` field in the request body, matching your deployment's configured model name
+
+:::note
+`served_model_name` is optional. If your deployment's model name is empty, you can omit it and just use `model: baseten/{deployment_id}`.
+:::

litellm/llms/baseten/chat.py

@@ -1,5 +1,7 @@
-from typing import Optional
+from typing import List, Optional
+
 from litellm.llms.openai.chat.gpt_transformation import OpenAIGPTConfig
+from litellm.types.llms.openai import AllMessageValues
 
 
 class BasetenConfig(OpenAIGPTConfig):
@@ -82,6 +84,28 @@ def map_openai_params(
                 optional_params[param] = value
         return optional_params
 
+    def transform_request(
+        self,
+        model: str,
+        messages: List[AllMessageValues],
+        optional_params: dict,
+        litellm_params: dict,
+        headers: dict,
+    ) -> dict:
+        # For dedicated deployments, the model is the deployment ID (e.g. "wd1lndkw")
+        # but the server may expect a different model name in the request body
+        served_model_name = litellm_params.get("served_model_name")
+        if served_model_name:
+            model = served_model_name
+
+        return super().transform_request(
+            model=model,
+            messages=messages,
+            optional_params=optional_params,
+            litellm_params=litellm_params,
+            headers=headers,
+        )
+
     def _get_openai_compatible_provider_info(
         self, api_base: str, api_key: str
     ) -> tuple:

tests/test_litellm/llms/baseten/chat/test_baseten_completions.py

@@ -50,5 +50,53 @@ def test_model_api_inference(self):
         assert api_key == "test-key"
 
 
+class TestBasetenTransformRequest:
+    """Test Baseten transform_request with served_model_name"""
+
+    def test_transform_request_with_served_model_name(self):
+        """Test that served_model_name overrides the model in request body"""
+        config = BasetenConfig()
+
+        result = config.transform_request(
+            model="wd1lndkw",
+            messages=[{"role": "user", "content": "Hello!"}],
+            optional_params={},
+            litellm_params={"served_model_name": "baseten-hosted/zai-org/GLM-5"},
+            headers={},
+        )
+
+        assert result["model"] == "baseten-hosted/zai-org/GLM-5"
+        assert result["messages"] == [{"role": "user", "content": "Hello!"}]
+
+    def test_transform_request_without_served_model_name(self):
+        """Test that model is used as-is when served_model_name is not set"""
+        config = BasetenConfig()
+
+        result = config.transform_request(
+            model="wd1lndkw",
+            messages=[{"role": "user", "content": "Hello!"}],
+            optional_params={},
+            litellm_params={},
+            headers={},
+        )
+
+        assert result["model"] == "wd1lndkw"
+
+    def test_transform_request_model_api_with_served_model_name(self):
+        """Test served_model_name also works for Model API models"""
+        config = BasetenConfig()
+
+        result = config.transform_request(
+            model="openai/gpt-oss-120b",
+            messages=[{"role": "user", "content": "Hello!"}],
+            optional_params={"max_tokens": 100},
+            litellm_params={"served_model_name": "my-custom-name"},
+            headers={},
+        )
+
+        assert result["model"] == "my-custom-name"
+        assert result["max_tokens"] == 100
+
+
 if __name__ == "__main__":
     pytest.main([__file__])