diff --git a/tests/entrypoints/openai/responses/test_harmony.py b/tests/entrypoints/openai/responses/test_harmony.py
index e99a299c12ee..fe67147ebc04 100644
--- a/tests/entrypoints/openai/responses/test_harmony.py
+++ b/tests/entrypoints/openai/responses/test_harmony.py
@@ -1259,27 +1259,28 @@ async def test_system_prompt_override(client: OpenAI, model_name: str):
     assert response.status == "completed"
     assert response.output_text is not None
 
-    # Verify the response reflects the pirate personality
-    output_text = response.output_text.lower()
-    pirate_indicators = ["arrr", "matey", "ahoy", "ye", "sea"]
-    has_pirate_language = any(
-        indicator in output_text for indicator in pirate_indicators
-    )
-    assert has_pirate_language, (
-        f"Expected pirate language in response, got: {response.output_text}"
-    )
-
-    # Verify the reasoning mentions the custom system prompt
+    # Extract reasoning first (needed for relaxed persona check)
     reasoning_item = None
     for item in response.output:
         if item.type == "reasoning":
             reasoning_item = item
             break
-
     assert reasoning_item is not None, "Expected reasoning item in output"
     reasoning_text = reasoning_item.content[0].text.lower()
-    assert "pirate" in reasoning_text, (
-        f"Expected reasoning to mention pirate, got: {reasoning_text}"
+
+    # Verify the custom system prompt was applied: either response uses pirate
+    # language, or reasoning mentions the pirate persona. Models may occasionally
+    # produce generic replies despite considering the persona.
+    output_text = response.output_text.lower()
+    pirate_indicators = ["arrr", "matey", "ahoy", "ye", "sea", "aye"]
+    has_pirate_language = any(
+        indicator in output_text for indicator in pirate_indicators
+    )
+    reasoning_mentions_pirate = "pirate" in reasoning_text
+    assert has_pirate_language or reasoning_mentions_pirate, (
+        f"Expected pirate language in response or 'pirate' in reasoning. "
+        f"Response: {response.output_text!r}. Reasoning excerpt: "
+        f"{reasoning_text[:200]!r}..."
     )
 
     # Test 2: Verify system message is not duplicated in input_messages