The 5-hour Codex cap returns for Plus accounts

Starting August 25, ChatGPT Plus subscribers once again have a five-hour usage window in Codex and ChatGPT Work. For several weeks before that, the weekly cap was the only limit. Thibault “Tibo” Sottiaux, OpenAI’s engineering lead for Codex and ChatGPT, announced the change on X the evening before, 9to5mac reported.
The window opens with your first message. It is shared across the local Codex CLI, Codex cloud tasks and ChatGPT Work, while regular ChatGPT chat has a separate allowance. The weekly cap still applies on top and runs on a seven-day cycle, so a Plus user needs room under both limits to keep working. If a limit hits in the middle of a turn, Codex may finish that turn within fair-use bounds.
What fills the window is tokens. Since April, Plus, Pro and Business have been metered directly in tokens instead of messages, so one noisy debugging turn can cost more than a dozen short prompts. That makes the “accidentally” in Sottiaux’s explanation easy to believe.
He gave two reasons. The five-hour limit “allows us to smoothen the load on our compute,” which in his telling keeps the plan “generous in terms of weekly usage.” The second reason is about who uses Plus. These are “relatively casual and new users” who “just accidentally eat through their whole weeks usage and then are confused, making it not a great experience.” He had announced the return earlier and then postponed it, he wrote.
While the window was off, OpenAI also reset weekly usage early several times to mark milestones, including another million active users across Codex and ChatGPT Work, which are now one platform. Those free resets still happen now and then. Users can also earn banked resets through some promotions and referral offers. Anyone who hits either limit can wait for the next cycle or buy more credits. Buying an instant reset starts a new week with your next request.
OpenAI gives ranges, not fixed numbers, for how much a Plus window holds. Later, as of September 29, the estimates per five-hour window were roughly 15-150 messages on GPT-6 Sol, 5-45 on GPT-6 Astra and 350-3,000 on GPT-6 Luna. A cloud task may use more than a local message.
Sottiaux said nothing about Enterprise or Edu accounts, which use a separate credit-based system. He said the five-hour limit would stay off for Pro $100 and Pro $200 subscriptions “for the upcoming months.” By September and October, OpenAI’s own page said its Pro plans at $100, $200 and $500 had no five-hour limit at all, though weekly limits may still apply.
Up next
GPT-5.6 becomes an option inside Kiro
Developers working inside Kiro can now pick GPT-5.6 to plan, build, review and test code, and OpenAI is pitching it on price-performance after cutting Sol’s API and credit pricing by over 20% on August 21, 2026.
Related stories
- ChatGPT and Codex will share one usage pool in December
- A $200 Codex weekly limit burned in a median 23 hours
- OpenAI doubles Pro and Business Premium usage limits
- Codex will spend a saved limit reset, but only when asked
- GPT-6.1 Sol gets 8x speed at $60 per million output tokens
- Codex reset hunters lose Day 1 to a 50% speedup
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.

