diff --git a/benchmarks/exercism_practice.py b/benchmarks/exercism_practice.py
index ddda659f9..1d1a4cd39 100755
--- a/benchmarks/exercism_practice.py
+++ b/benchmarks/exercism_practice.py
@@ -64,7 +64,7 @@ async def failure_analysis(exercise_runner, language):
     try:
         llm_api_handler = SESSION_CONTEXT.get().llm_api_handler
         llm_grade = await llm_api_handler.call_llm_api(messages, model, False)
-        response = llm_grade.choices[0].message.content
+        response = llm_grade.text
     except BadRequestError:
         response = "Unable to analyze test case\nreason: too many tokens to analyze"