diff --git a/evaluation/benchmarks/swe_bench/run_infer.py b/evaluation/benchmarks/swe_bench/run_infer.py
index ca9dd52ec3..30a40c8e66 100644
--- a/evaluation/benchmarks/swe_bench/run_infer.py
+++ b/evaluation/benchmarks/swe_bench/run_infer.py
@@ -109,9 +109,7 @@ def get_instruction(instance: pd.Series, metadata: EvalMetadata) -> MessageActio
         template_name = 'swt.j2'
     elif mode == 'swe':
         if 'claude' in llm_model:
-            template_name = 'swe_claude.j2'
-        elif 'gemini' in llm_model:
-            template_name = 'swe_gemini.j2'
+            template_name = 'swe_default.j2'
         elif 'gpt-4.1' in llm_model:
             template_name = 'swe_gpt4.j2'
         else: