diff --git a/litellm/proxy/auth/model_checks.py b/litellm/proxy/auth/model_checks.py
index 55b4670d14e..fb925a3b133 100644
--- a/litellm/proxy/auth/model_checks.py
+++ b/litellm/proxy/auth/model_checks.py
@@ -325,7 +325,7 @@ def get_known_models_from_wildcard(wildcard_model: str, litellm_params: LiteLLM_
             # Only strip the leading segment when it is a known provider, so ids whose first
             # segment is an org rather than a provider (e.g. "meta-llama/Llama-3-8B") keep it.
             leading, sep, model_suffix = model.partition("/")
-            if sep and leading in known_providers:
+            if sep and leading in known_providers and (provider in litellm.models_by_provider or leading == provider):
                 model = f"{wildcard_provider_prefix}/{model_suffix}"
             else:
                 model = f"{wildcard_provider_prefix}/{model}"
diff --git a/tests/test_litellm/proxy/auth/test_model_checks.py b/tests/test_litellm/proxy/auth/test_model_checks.py
index 451f4eac9a0..bea46a3fd23 100644
--- a/tests/test_litellm/proxy/auth/test_model_checks.py
+++ b/tests/test_litellm/proxy/auth/test_model_checks.py
@@ -442,6 +442,7 @@ def test_get_known_models_from_wildcard_hosted_vllm_uses_provider_endpoint():
         "data": [
             {"id": "meta-llama/Llama-3.1-8B-Instruct"},
             {"id": "qwen2.5"},
+            {"id": "openai/foo"},
         ]
     }
     original_check_provider_endpoint = litellm.check_provider_endpoint
@@ -460,6 +461,7 @@ def test_get_known_models_from_wildcard_hosted_vllm_uses_provider_endpoint():
     assert result == [
         "hosted_vllm/meta-llama/Llama-3.1-8B-Instruct",
         "hosted_vllm/qwen2.5",
+        "hosted_vllm/openai/foo",
     ]