GPT-5.1 Codex-Max Works All Day

Today we at @OpenAI are releasing GPT-5.1-Codex-Max, which can work autonomously for more than a day over millions of tokens. Pretraining hasn't hit a wall, and neither has test-time compute. Congrats to my teammates @kevinleestone & @mikegmalek for helping to make it possible! Running tasks ... enables GPT-5.1-Codex-Max to complete tasks that previously failed due to context-window limits, such as coding and long-running agent loops by pruning its history and keeping the most important context over long horizons. In Codex sessions, GPT-5.1-Codex-Max automatically compacts its session as it approaches its context window limit, giving it a fresh context. It repeats this process until the task is completed. ... to sustain coherent work over long horizons is a foundation on the path toward more general, reliable AI systems ... can work independently for hours at a time. In our interactions, we've observed GPT-5.1-Codex-Max work on tasks for hours. It will persistently iterate on its implementation, and ultimately deliver a successful result. ...ch Verified (n=500) GPT-5.1-Codex GPT-5.1-Codex-Max high high medium medium low low Thinking tokens

A dark-mode announcement claims that GPT-5.1-Codex-Max can work autonomously for more than a day and sustain long-running coding tasks. The lower panels show task-running explanations and a chart where Codex-Max uses more thinking tokens and performs better across low, medium, and high difficulty. The joke is a sarcastic reaction to the grand announcement, implying that impressive autonomous operation is unnecessary if nobody needs an agent running continuously for more than a day.

• @dev_meme / 7445 2025-11-19