diff --git a/litellm/proxy/management_endpoints/model_management_endpoints.py b/litellm/proxy/management_endpoints/model_management_endpoints.py
index d8a1075a16..721a8c2a1c 100644
--- a/litellm/proxy/management_endpoints/model_management_endpoints.py
+++ b/litellm/proxy/management_endpoints/model_management_endpoints.py
@@ -521,6 +521,16 @@ async def _update_existing_team_model_assignment(
                 and _get_team_public_model_name(d.model_info) == old_public_name
             ]
 
+        # Add new name first, then delete old name to prevent access loss on partial failure
+        await team_model_add(
+            data=TeamModelAddRequest(
+                team_id=team_id,
+                models=[public_model_name],
+            ),
+            http_request=Request(scope={"type": "http"}),
+            user_api_key_dict=user_api_key_dict,
+        )
+
         if not other_deployments_with_old_name:
             await team_model_delete(
                 data=TeamModelDeleteRequest(
@@ -531,15 +541,6 @@ async def _update_existing_team_model_assignment(
                 user_api_key_dict=user_api_key_dict,
             )
 
-        await team_model_add(
-            data=TeamModelAddRequest(
-                team_id=team_id,
-                models=[public_model_name],
-            ),
-            http_request=Request(scope={"type": "http"}),
-            user_api_key_dict=user_api_key_dict,
-        )
-
     patch_data.model_name = None
 
 
diff --git a/litellm/router.py b/litellm/router.py
index 76c13443e7..d7f5d42eac 100644
--- a/litellm/router.py
+++ b/litellm/router.py
@@ -8233,9 +8233,11 @@ class Router:
         ):
             return True
         elif model_name is not None and model["model_name"] == model_name:
+            model_team_id = (model.get("model_info") or {}).get("team_id")
             if (
                 team_id is None
-                or (model.get("model_info") or {}).get("team_id") == team_id
+                or model_team_id is None  # global deployment - accessible to all teams
+                or model_team_id == team_id
             ):
                 return True
         return False
diff --git a/tests/test_litellm/proxy/management_endpoints/test_model_management_endpoints.py b/tests/test_litellm/proxy/management_endpoints/test_model_management_endpoints.py
index 83e6b0c93a..2dd29fd5c9 100644
--- a/tests/test_litellm/proxy/management_endpoints/test_model_management_endpoints.py
+++ b/tests/test_litellm/proxy/management_endpoints/test_model_management_endpoints.py
@@ -712,6 +712,15 @@ class TestTeamModelSiblingRouting:
                         "team_public_model_name": public_name,
                     },
                 },
+                {
+                    "model_name": "global-gpt-4o",
+                    "litellm_params": {
+                        "model": "azure/gpt-4o",
+                        "api_key": "global-key",
+                        "api_base": "https://global.openai.azure.com",
+                    },
+                    "model_info": {},  # No team_id - global deployment
+                },
             ],
         )
 
@@ -732,6 +741,38 @@ class TestTeamModelSiblingRouting:
             "https://westus.openai.azure.com",
         }
 
+    def test_global_deployments_accessible_to_teams(self):
+        """Test that global deployments (no team_id) are accessible to all teams"""
+        import litellm
+
+        router = litellm.Router(
+            model_list=[
+                {
+                    "model_name": "global-gpt-4o",
+                    "litellm_params": {
+                        "model": "azure/gpt-4o",
+                        "api_key": "global-key",
+                        "api_base": "https://global.openai.azure.com",
+                    },
+                    "model_info": {},  # No team_id - global deployment
+                },
+            ],
+        )
+
+        # Global deployment should be accessible when team_id is provided
+        deployments = router._get_all_deployments(
+            model_name="global-gpt-4o", team_id="teamA"
+        )
+        assert len(deployments) == 1
+        assert deployments[0]["model_name"] == "global-gpt-4o"
+
+        # should_include_deployment should return True for global deployments
+        assert router.should_include_deployment(
+            model_name="global-gpt-4o",
+            model={"model_name": "global-gpt-4o", "model_info": {}},
+            team_id="teamA",
+        )
+
 
 class TestTeamModelUpdate:
     """Test team model update handles team_id consistently with model creation"""