diff --git a/src/llm_core.py b/src/llm_core.py
index b012638fa..8da2c46e0 100644
--- a/src/llm_core.py
+++ b/src/llm_core.py
@@ -563,8 +563,9 @@ def _build_chatgpt_responses_payload(
     }
     if not _restricts_temperature(model):
         payload["temperature"] = temperature
-    if max_tokens and max_tokens > 0:
-        payload["max_output_tokens"] = max_tokens
+    # ChatGPT Subscription Codex API does not support max_output_tokens —
+    # passing it returns HTTP 400 "Unsupported parameter: max_output_tokens".
+    # Do not include it in the payload.
     return payload
 
 
diff --git a/tests/test_llm_core_temperature.py b/tests/test_llm_core_temperature.py
index f49d3dba0..121a7ff4b 100644
--- a/tests/test_llm_core_temperature.py
+++ b/tests/test_llm_core_temperature.py
@@ -75,7 +75,10 @@ def test_normal_model_payload_keeps_temperature_above_one(monkeypatch):
     assert payload["temperature"] == 1.2
 
 
-def test_chatgpt_subscription_payload_uses_max_output_tokens():
+def test_chatgpt_subscription_payload_omits_max_output_tokens():
+    # ChatGPT Subscription Codex API does not support max_output_tokens —
+    # passing it returns HTTP 400 "Unsupported parameter: max_output_tokens".
+    # The payload should NOT include max_output_tokens regardless of max_tokens.
     payload = llm_core._build_chatgpt_responses_payload(
         "gpt-5.1-codex",
         [{"role": "user", "content": "Say OK"}],
@@ -83,10 +86,10 @@ def test_chatgpt_subscription_payload_uses_max_output_tokens():
         max_tokens=37,
     )
 
-    assert payload["max_output_tokens"] == 37
+    assert "max_output_tokens" not in payload
 
 
-def test_chatgpt_subscription_payload_omits_empty_max_output_tokens():
+def test_chatgpt_subscription_payload_omits_max_output_tokens_when_zero():
     payload = llm_core._build_chatgpt_responses_payload(
         "gpt-5.1-codex",
         [{"role": "user", "content": "Say OK"}],