diff --git a/backend/open_webui/models/chat_messages.py b/backend/open_webui/models/chat_messages.py
index 4baa7ea20..e0f00da56 100644
--- a/backend/open_webui/models/chat_messages.py
+++ b/backend/open_webui/models/chat_messages.py
@@ -140,9 +140,11 @@ class ChatMessageTable:
                     )
                 if "error" in data:
                     existing.error = data.get("error")
-                # Extract usage from info.usage if present
-                info = data.get("info", {})
-                usage = info.get("usage") if info else None
+                # Extract usage - check direct field first, then info.usage
+                usage = data.get("usage")
+                if not usage:
+                    info = data.get("info", {})
+                    usage = info.get("usage") if info else None
                 if usage:
                     existing.usage = usage
                 existing.updated_at = now
@@ -151,9 +153,11 @@ class ChatMessageTable:
                 return ChatMessageModel.model_validate(existing)
             else:
                 # Insert new
-                # Extract usage from info.usage if present
-                info = data.get("info", {})
-                usage = info.get("usage") if info else None
+                # Extract usage - check direct field first, then info.usage
+                usage = data.get("usage")
+                if not usage:
+                    info = data.get("info", {})
+                    usage = info.get("usage") if info else None
                 message = ChatMessage(
                     id=message_id,
                     chat_id=chat_id,
diff --git a/backend/open_webui/utils/middleware.py b/backend/open_webui/utils/middleware.py
index e4d800ba5..cf0f14501 100644
--- a/backend/open_webui/utils/middleware.py
+++ b/backend/open_webui/utils/middleware.py
@@ -3042,6 +3042,7 @@ async def process_chat_response(
                 async def stream_body_handler(response, form_data):
                     nonlocal content
                     nonlocal content_blocks
+                    nonlocal usage
 
                     response_tool_calls = []