diff --git a/webapp/api/score.py b/webapp/api/score.py
index 65f2ea9..4f2a303 100644
--- a/webapp/api/score.py
+++ b/webapp/api/score.py
@@ -90,6 +90,10 @@ def score_sample(
     - `factual_correctness` — 回答与参考答案的事实准确性（需 ground_truth）
     - `semantic_similarity` — 回答与参考答案的语义相似度（需 ground_truth）
 
+    **推荐模型配置**：
+    - `judge_model`: `gpt-5.4`
+    - `embedding_model`: `text-embedding-3-small`
+
     **鉴权**：若 `.env` 中配置了 `SCORE_API_TOKEN`，需在请求头携带
     `Authorization: Bearer <token>`；留空则无需鉴权（适合内网部署）。
     """
diff --git a/webapp/models.py b/webapp/models.py
index 69f5010..71a6801 100644
--- a/webapp/models.py
+++ b/webapp/models.py
@@ -420,8 +420,8 @@ class ScoreRequest(BaseModel):
                     "context_recall",
                     "context_precision",
                 ],
-                "judge_model": "deepseek-v4-flash",
-                "embedding_model": "text-embedding-v3",
+                "judge_model": "gpt-5.4",
+                "embedding_model": "text-embedding-3-small",
             }
         }
     )