fix(router): use custom model_id pricing when litellm_model_name_model_info is missing

Return custom_model_info from get_deployment_model_info when
litellm_model_name_model_info is unavailable. This preserves explicit
input_cost_per_token and output_cost_per_token values (including 0.0)
for custom deployments and prevents model_group_info from defaulting to
None costs.

litellm/router.py

@@ -7754,6 +7754,8 @@ class Router:
             )
         elif litellm_model_name_model_info is not None:
             model_info = litellm_model_name_model_info
+        elif custom_model_info is not None:
+            model_info = cast(ModelInfo, custom_model_info)
 
         return model_info
 

tests/test_litellm/test_router.py

@@ -1955,6 +1955,25 @@ def test_get_deployment_model_info_base_model_flow():
             # Should return None when no model info is found
             assert result is None
 
+    # Test Case 6: Only custom model info exists (no model-name info)
+    mock_custom_only_info = {
+        "input_cost_per_token": 0.0,
+        "output_cost_per_token": 0.0,
+        "mode": "chat",
+        "litellm_provider": "openai",
+    }
+
+    with patch.object(litellm, "model_cost", {"custom-only-id": mock_custom_only_info}):
+        with patch.object(litellm, "get_model_info", return_value=None):
+            result = router.get_deployment_model_info(
+                model_id="custom-only-id", model_name="missing-model-name"
+            )
+
+            assert result is not None
+            assert result["input_cost_per_token"] == 0.0
+            assert result["output_cost_per_token"] == 0.0
+            assert result["litellm_provider"] == "openai"
+
     print("✓ All base model flow test cases passed!")