diff --git a/backend/open_webui/main.py b/backend/open_webui/main.py
index 914c0620b8..36043d071a 100644
--- a/backend/open_webui/main.py
+++ b/backend/open_webui/main.py
@@ -587,7 +587,7 @@ app.state.SCIM_TOKEN = SCIM_TOKEN
 ########################################
 
 app.state.BASE_MODELS = []
-app.state.BASE_MODELS_EPOCH = None
+app.state.BASE_MODELS_EPOCH = ''
 
 ########################################
 #
diff --git a/backend/open_webui/routers/openai.py b/backend/open_webui/routers/openai.py
index 7ea9302d47..fc6a6ee180 100644
--- a/backend/open_webui/routers/openai.py
+++ b/backend/open_webui/routers/openai.py
@@ -345,7 +345,6 @@ async def get_openai_connection(idx: int) -> tuple[str, str, dict]:
 
 
 async def clear_openai_model_cache(request: Request):
-    await Config.upsert({'models.base_models_cache_epoch': str(uuid4())})
     await get_all_models.cache.clear()
     request.app.state.BASE_MODELS = []
     request.app.state.OPENAI_MODELS = {}
@@ -570,10 +569,18 @@ async def update_config(request: Request, form_data: OpenAIConfigForm, user=Depe
             'openai.api_base_urls': form_data.OPENAI_API_BASE_URLS,
             'openai.api_keys': api_keys,
             'openai.api_configs': api_configs,
+            'models.base_models_cache_epoch': str(uuid4()),
         }
     )
 
-    await clear_openai_model_cache(request)
+    await get_all_models.cache.clear()
+    request.app.state.BASE_MODELS = []
+    request.app.state.OPENAI_MODELS = {}
+    models = getattr(request.app.state, 'MODELS', None)
+    if hasattr(models, 'clear'):
+        models.clear()
+    else:
+        request.app.state.MODELS = {}
 
     await publish_event(
         request,
diff --git a/backend/open_webui/utils/models.py b/backend/open_webui/utils/models.py
index d55c5b52bc..89e104a869 100644
--- a/backend/open_webui/utils/models.py
+++ b/backend/open_webui/utils/models.py
@@ -72,25 +72,26 @@ async def get_all_models(request, refresh: bool = False, user: UserModel = None)
         'evaluation.arena.models',
         'models.default_metadata',
     )
-    # A changed epoch means another instance edited a provider, so our cached models are stale.
     cache_epoch = config.get('models.base_models_cache_epoch')
-    stale = request.app.state.BASE_MODELS_EPOCH != cache_epoch
+    if request.app.state.BASE_MODELS_EPOCH != cache_epoch:
+        # Another instance changed a provider connection, so our cached models are outdated.
+        refresh = True
 
-    if refresh or stale:
+    if refresh:
         await openai.get_all_models.cache.clear()
         await ollama.get_all_models.cache.clear()
 
     if (
         request.app.state.MODELS
         and request.app.state.BASE_MODELS
-        and (config.get('models.base_models_cache') and not (refresh or stale))
+        and (config.get('models.base_models_cache') and not refresh)
     ):
         base_models = request.app.state.BASE_MODELS
     else:
         base_models = await get_all_base_models(request, user=user)
-        request.app.state.BASE_MODELS_EPOCH = cache_epoch
         if base_models:
             request.app.state.BASE_MODELS = base_models
+            request.app.state.BASE_MODELS_EPOCH = cache_epoch
         else:
             base_models = request.app.state.BASE_MODELS