diff --git a/llama_stack/providers/inline/inference/vllm/vllm.py b/llama_stack/providers/inline/inference/vllm/vllm.py
index 73f7adecd8..a2df16ed2a 100644
--- a/llama_stack/providers/inline/inference/vllm/vllm.py
+++ b/llama_stack/providers/inline/inference/vllm/vllm.py
@@ -67,7 +67,7 @@ def __init__(self, config: VLLMConfig):
         self.formatter = ChatFormat(Tokenizer.get_instance())
 
     async def initialize(self):
-        log.info("Initializing vLLM inference adapter")
+        log.info("Initializing vLLM inference provider.")
 
         # Disable usage stats reporting. This would be a surprising thing for most
         # people to find out was on by default.
@@ -95,15 +95,36 @@ async def initialize(self):
         self.engine = AsyncLLMEngine.from_engine_args(engine_args)
 
     async def shutdown(self):
-        """Shutdown the vLLM inference adapter."""
-        log.info("Shutting down vLLM inference adapter")
+        """Shut down the vLLM inference adapter."""
+        log.info("Shutting down vLLM inference provider.")
         if self.engine:
             self.engine.shutdown_background_loop()
 
-    async def register_model(self, model: Model) -> None:
-        raise ValueError(
-            "You cannot dynamically add a model to a running vllm instance"
-        )
+    # Note that the return type of the superclass method is WRONG
+    async def register_model(self, model: Model) -> Model:
+        """
+        Callback that is called when the server associates an inference endpoint
+        with an inference provider.
+
+        :param model: Object that encapsulates parameters necessary for identifying
+         a specific LLM.
+
+        :returns: The input ``Model`` object. It may or may not be permissible
+         to change fields before returning this object.
+        """
+        log.info(f"Registering model {model.identifier} with vLLM inference provider.")
+        # The current version of this provided is hard-coded to serve only
+        # the model specified in the YAML config file.
+        configured_model = resolve_model(self.config.model)
+        registered_model = resolve_model(model.model_id)
+
+        if configured_model.core_model_id != registered_model.core_model_id:
+            raise ValueError(
+                f"Requested model '{model.identifier}' is different from "
+                f"model '{self.config.model}' that this provider "
+                f"is configured to serve"
+            )
+        return model
 
     def _sampling_params(self, sampling_params: SamplingParams) -> VLLMSamplingParams:
         if sampling_params is None:
@@ -167,7 +188,9 @@ async def chat_completion(
         log.info("Sampling params: %s", sampling_params)
         request_id = _random_uuid()
 
-        prompt = await chat_completion_request_to_prompt(request, self.formatter)
+        prompt = await chat_completion_request_to_prompt(
+            request, self.config.model, self.formatter
+        )
         vllm_sampling_params = self._sampling_params(request.sampling_params)
         results_generator = self.engine.generate(
             prompt, vllm_sampling_params, request_id
diff --git a/llama_stack/providers/tests/inference/fixtures.py b/llama_stack/providers/tests/inference/fixtures.py
index d956caa930..b6653b65d7 100644
--- a/llama_stack/providers/tests/inference/fixtures.py
+++ b/llama_stack/providers/tests/inference/fixtures.py
@@ -15,6 +15,7 @@
 from llama_stack.providers.inline.inference.meta_reference import (
     MetaReferenceInferenceConfig,
 )
+from llama_stack.providers.inline.inference.vllm import VLLMConfig
 from llama_stack.providers.remote.inference.bedrock import BedrockConfig
 
 from llama_stack.providers.remote.inference.cerebras import CerebrasImplConfig
@@ -105,6 +106,26 @@ def inference_ollama(inference_model) -> ProviderFixture:
     )
 
 
+@pytest_asyncio.fixture(scope="session")
+def inference_vllm(inference_model) -> ProviderFixture:
+    inference_model = (
+        [inference_model] if isinstance(inference_model, str) else inference_model
+    )
+    return ProviderFixture(
+        providers=[
+            Provider(
+                provider_id=f"vllm-{i}",
+                provider_type="inline::vllm",
+                config=VLLMConfig(
+                    model=m,
+                    enforce_eager=True,  # Make test run faster
+                ).model_dump(),
+            )
+            for i, m in enumerate(inference_model)
+        ]
+    )
+
+
 @pytest.fixture(scope="session")
 def inference_vllm_remote() -> ProviderFixture:
     return ProviderFixture(
@@ -253,6 +274,7 @@ def model_id(inference_model) -> str:
     "ollama",
     "fireworks",
     "together",
+    "vllm",
     "groq",
     "vllm_remote",
     "remote",
@@ -286,4 +308,8 @@ async def inference_stack(request, inference_model):
         ],
     )
 
-    return test_stack.impls[Api.inference], test_stack.impls[Api.models]
+    # Pytest yield fixture; see https://docs.pytest.org/en/stable/how-to/fixtures.html#yield-fixtures-recommended
+    yield test_stack.impls[Api.inference], test_stack.impls[Api.models]
+
+    # Cleanup code that runs after test case completion
+    await test_stack.impls[Api.inference].shutdown()
diff --git a/llama_stack/providers/tests/inference/test_text_inference.py b/llama_stack/providers/tests/inference/test_text_inference.py
index 7776c79598..e2c9399149 100644
--- a/llama_stack/providers/tests/inference/test_text_inference.py
+++ b/llama_stack/providers/tests/inference/test_text_inference.py
@@ -86,7 +86,9 @@ def sample_tool_definition():
 
 
 class TestInference:
-    @pytest.mark.asyncio
+    # Session scope for asyncio because the tests in this class all
+    # share the same provider instance.
+    @pytest.mark.asyncio(loop_scope="session")
     async def test_model_list(self, inference_model, inference_stack):
         _, models_impl = inference_stack
         response = await models_impl.list_models()
@@ -102,7 +104,7 @@ async def test_model_list(self, inference_model, inference_stack):
 
         assert model_def is not None
 
-    @pytest.mark.asyncio
+    @pytest.mark.asyncio(loop_scope="session")
     async def test_completion(self, inference_model, inference_stack):
         inference_impl, _ = inference_stack
 
@@ -147,7 +149,7 @@ async def test_completion(self, inference_model, inference_stack):
         last = chunks[-1]
         assert last.stop_reason == StopReason.out_of_tokens
 
-    @pytest.mark.asyncio
+    @pytest.mark.asyncio(loop_scope="session")
     async def test_completion_logprobs(self, inference_model, inference_stack):
         inference_impl, _ = inference_stack
 
@@ -202,7 +204,7 @@ async def test_completion_logprobs(self, inference_model, inference_stack):
             else:  # no token, no logprobs
                 assert not chunk.logprobs, "Logprobs should be empty"
 
-    @pytest.mark.asyncio
+    @pytest.mark.asyncio(loop_scope="session")
     @pytest.mark.skip("This test is not quite robust")
     async def test_completion_structured_output(self, inference_model, inference_stack):
         inference_impl, _ = inference_stack
@@ -247,7 +249,7 @@ class Output(BaseModel):
         assert answer.year_born == "1963"
         assert answer.year_retired == "2003"
 
-    @pytest.mark.asyncio
+    @pytest.mark.asyncio(loop_scope="session")
     async def test_chat_completion_non_streaming(
         self, inference_model, inference_stack, common_params, sample_messages
     ):
@@ -264,7 +266,7 @@ async def test_chat_completion_non_streaming(
         assert isinstance(response.completion_message.content, str)
         assert len(response.completion_message.content) > 0
 
-    @pytest.mark.asyncio
+    @pytest.mark.asyncio(loop_scope="session")
     async def test_structured_output(
         self, inference_model, inference_stack, common_params
     ):
@@ -335,7 +337,7 @@ class AnswerFormat(BaseModel):
         with pytest.raises(ValidationError):
             AnswerFormat.model_validate_json(response.completion_message.content)
 
-    @pytest.mark.asyncio
+    @pytest.mark.asyncio(loop_scope="session")
     async def test_chat_completion_streaming(
         self, inference_model, inference_stack, common_params, sample_messages
     ):
@@ -362,7 +364,7 @@ async def test_chat_completion_streaming(
         end = grouped[ChatCompletionResponseEventType.complete][0]
         assert end.event.stop_reason == StopReason.end_of_turn
 
-    @pytest.mark.asyncio
+    @pytest.mark.asyncio(loop_scope="session")
     async def test_chat_completion_with_tool_calling(
         self,
         inference_model,
@@ -409,7 +411,7 @@ async def test_chat_completion_with_tool_calling(
         assert "location" in call.arguments
         assert "San Francisco" in call.arguments["location"]
 
-    @pytest.mark.asyncio
+    @pytest.mark.asyncio(loop_scope="session")
     async def test_chat_completion_with_tool_calling_streaming(
         self,
         inference_model,