diff --git a/src/bub/builtin/codex_provider.py b/src/bub/builtin/codex_provider.py
index e9eddbeb..2aa523fb 100644
--- a/src/bub/builtin/codex_provider.py
+++ b/src/bub/builtin/codex_provider.py
@@ -136,9 +136,6 @@ def _completion_params_to_responses_params(self, params: CompletionParams) -> Re
             response_format=params.response_format,
             stream=params.stream,
             parallel_tool_calls=params.parallel_tool_calls,
-            max_output_tokens=(
-                params.max_completion_tokens if params.max_completion_tokens is not None else params.max_tokens
-            ),
             reasoning=reasoning,
         )
 
diff --git a/tests/test_builtin_codex.py b/tests/test_builtin_codex.py
index 98e44f11..98d122b9 100644
--- a/tests/test_builtin_codex.py
+++ b/tests/test_builtin_codex.py
@@ -189,25 +189,10 @@ def test_codex_completion_params_use_official_responses_payload_fields() -> None
     assert payload == {
         "model": "gpt-5.5",
         "input": [{"role": "user", "content": "hello"}],
-        "max_output_tokens": 100,
         "stream": True,
     }
 
 
-def test_codex_completion_params_prefer_max_completion_tokens() -> None:
-    provider = OpenaiCodexProvider(api_key=_jwt_with_account("acct_123"))
-    params = CompletionParams(
-        model_id="gpt-5.5",
-        messages=[{"role": "user", "content": "hello"}],
-        max_tokens=100,
-        max_completion_tokens=200,
-    )
-
-    responses_params = provider._completion_params_to_responses_params(params)
-
-    assert responses_params.max_output_tokens == 200
-
-
 def test_codex_completion_params_convert_chat_tool_messages_to_responses_items() -> None:
     provider = OpenaiCodexProvider(api_key=_jwt_with_account("acct_123"))
     params = CompletionParams(