Home / Guides / Codex Rate Limits
Codex 5-Hour Limit: Back on Plus, Still Off on Pro
Yes, Codex has a five-hour limit again. OpenAI suspended it on 12 July 2026 and restored it for Plus on 25 August 2026. Plus and Standard Business have it; Pro plans currently do not. Weekly limits can apply on top, and whichever you reach first limits further requests. An active turn can continue, subject to fair use limits.
Rather not rerun /status? A Codex reset tracker keeps both reset times in your menu bar or taskbar.
Limits checked 3 October 2026 against OpenAI's Codex pricing page · By Soren Starck
On Windows? Track the Codex limit on Windows: a native app for Windows 10 and 11, with alerts before you hit a limit.
How much fits in five hours
OpenAI does not publish a fixed number. It publishes ranges of local messages per five-hour period, per model, and says local messages and cloud chats share the same allowance:
| Model | Plus, per 5 hours | Standard Business, per 5 hours |
|---|---|---|
| GPT-6 Astra | 5 to 45 messages | 5 to 45 messages |
| GPT-6.1 Sol | 15 to 160 messages | 15 to 160 messages |
| GPT-6 Sol | 15 to 150 messages | 15 to 150 messages |
| GPT-6 Luna | 350 to 3,000 messages | 350 to 3,000 messages |
The ranges are wide because a message is not a fixed unit. OpenAI's own words: small scripts or routine functions may use only a fraction of your allowance, while larger projects, long-running tasks or sessions that hold more context use significantly more per message. Source: OpenAI Codex pricing, checked 3 October 2026.
What changed, and when
- 12 July 2026: OpenAI temporarily suspended the Codex five-hour limit for Plus, Business and Pro. Only the weekly cap was left.
- 25 August 2026: the five-hour limit came back for Plus, announced the day before by Codex lead Thibault Sottiaux (reported by 9to5Mac).
- Pro ($100 and $200): OpenAI said the five-hour limit stays disabled for the upcoming months, and its pricing page says Pro plans currently have none.
If you read in July or August that the five-hour limit was gone, that advice is out of date for Plus. It is still right for Pro, for now. Hitting the Plus limit most days and wondering whether Pro pays for itself? See is ChatGPT Pro worth it for Codex.
How the two limits stack
On Plus and Standard Business you have two limits at once: the five-hour one and, where it applies, a weekly one. A heavy morning can hit the five-hour wall with most of the week still available; a few heavy days in a row can hit the weekly wall while the five-hour limit is fine. Whichever you reach first limits further requests, and the reset you are waiting for is that limit's, not the other's. An active turn can continue, subject to fair use limits.
When you hit one, you can wait for it to reset or buy additional credits. To see which limit you hit and when it resets, open chatgpt.com/codex/settings/usage or run /status, or keep both reset times in view with a Codex reset tracker. If neither shows a limit and Codex still fails, check whether Codex is down right now.
The hard part is that both are invisible until you ask. That is why limits feel like outages: nothing warns you, and the first sign is Codex refusing work mid-task. If you are deciding whether a manual command, the usage page or a menu-bar app fits your workflow, compare the best Codex usage trackers before you buy.
SessionWatcher
Codex gives zero warning.
SessionWatcher gives all of them.
Native app for the Mac menu bar and the Windows taskbar. Track Claude, Codex and 18 more tools' usage, costs, and rate limits in real time.
“Fast, simple, and does exactly what it should. Definitely worth it.”
@nicojerome on GitHub
macOS 14+. Free for 7 days, no credit card.
How to keep both limits in view
SessionWatcher for Codex is a native app for the Mac menu bar and the Windows taskbar. It shows:
- Each limit OpenAI reports, five-hour and weekly, as a percentage used
- Reset times for each, so you know how long the wait is
- Local token totals and an estimated API-equivalent cost
- Desktop notifications before you reach a limit
Tips for staying under the Codex limits
- Watch both limits, not one. A healthy five-hour reading says nothing about the week.
- Match the model to the task. OpenAI's ranges differ by an order of magnitude between models; routine work does not need the heaviest one.
- Keep context tight. Long sessions that carry more context cost more per message.
- Plan heavy work around resets. If the five-hour limit is close, a break often costs less than hitting it mid-task.
Frequently Asked Questions
Does Codex have a 5-hour limit?
Yes, on Plus and Standard Business. OpenAI suspended the five-hour limit on 12 July 2026 and restored it for Plus on 25 August 2026. Its Codex pricing page (checked 3 October 2026) lists five-hour message estimates for Plus and Standard Business and says Pro plans currently have no five-hour limit. Weekly limits may apply on top.
Does Codex Pro have a 5-hour limit?
Not currently. When the limit came back for Plus, OpenAI said it would stay disabled for the $100 and $200 Pro plans for the upcoming months, and the pricing page says Pro plans currently have no five-hour limit. Weekly limits can still apply.
How many messages do I get in 5 hours on Codex Plus?
OpenAI publishes ranges rather than a fixed number, per model: 5 to 45 local messages for GPT-6 Astra, 15 to 160 for GPT-6.1 Sol, 15 to 150 for GPT-6 Sol and 350 to 3,000 for GPT-6 Luna, the same for Standard Business. Larger projects and long-running tasks use far more per message than small scripts.
When does my Codex limit reset?
It depends which limit stopped you. The five-hour limit covers a five-hour period, so it clears within hours; the weekly limit takes longer. Your exact reset time for each is on chatgpt.com/codex/settings/usage and in /status inside Codex CLI.
What happens when I hit a Codex limit?
Codex stops accepting requests until that limit resets, but an active turn can continue, subject to fair use limits. You can wait for the reset or buy additional credits. Whichever limit you reach first, five-hour or weekly, is the one that limits further requests.