diff --git a/src/libs/agent-runtime/huggingface/index.ts b/src/libs/agent-runtime/huggingface/index.ts
index b31a12b1af276..dab9dafbe7f40 100644
--- a/src/libs/agent-runtime/huggingface/index.ts
+++ b/src/libs/agent-runtime/huggingface/index.ts
@@ -23,7 +23,9 @@ export const LobeHuggingFaceAI = LobeOpenAICompatibleFactory({
     createChatCompletionStream: (client: HfInference, payload, instance) => {
       const { max_tokens = 4096} = payload;
       const hfRes = client.chatCompletionStream({
-        endpointUrl: instance.baseURL,
+        endpointUrl: instance.baseURL
+          ? new URL(payload.model, instance.baseURL).toString()
+          : instance.baseURL,
         max_tokens: max_tokens,
         messages: payload.messages,
         model: payload.model,