GPT-5.5 Codex reasoning-token clustering may be leading to degraded performance
Why CEREBRO kept it
GPT-5.5 Codex token-clustering bug; LLM token optimization mechanics.
The text below is an automated extraction of the article at https://github.com/openai/codex/issues/30364, stored verbatim in the public cerebro-vault repository. Copyright remains with the original publisher (github.com).
Summary I found an aggregate pattern in Codex token_count metadata: gpt-5.5 responses disproportionately land at exactly reasoning_output_tokens = 516, with additional fixed-boundary spikes around 1034 and 1552. This appears model-specific and coincides with lower overall reasoning-token intensity, which may help explain degraded performance on complex/high-stakes Codex tasks. This is related to #29353, which reported a task-level reproduction where gpt-5.5 runs ending at exactly 516 reasoning tokens returned the wrong answer. This issue adds aggregate evidence across a larger Feb-Jun window.
Community take
GPT-5.5 Codex caps reasoning tokens at multiples of 518 as a throughput optimization, causing predictable failures on complex tasks; users have switched to Claude.
Backlinks
Appeared in 1 briefing