diff --git a/server/inference.py b/server/inference.py
index 1d3a92b..4327aac 100644
--- a/server/inference.py
+++ b/server/inference.py
@@ -24,7 +24,7 @@ async def connect_to_conversation(ws: WebSocket):
   continuing = bool(ws.query_params.get("continue"))
   conversation_id = ws.path_params["conversation"]
   if conversation_id == "new":
-    conversation = llm.AsyncConversation(llm.get_async_model())
+    conversation = llm.AsyncConversation(llm.get_async_model(ws.query_params.get("model")))
   else:
     try:
       conversation: llm.AsyncConversation = llm.cli.load_conversation(conversation_id, async_=True)