Day 22/ Terra is back! […] gives 2X more usage than 6.1 Sol across API and subscriptions.
https://x.com/thsottiaux/status/2114800277677092936/openai — Codex usage limit resets
LAST RESET
2026-10-26
A change to the limit rules themselves, affecting your allowance going forward.
73
Total records
10
Banked resets
5.9d
Avg. interval
67.7d
Longest wait
SINCE LAST RESET1.9d
Average gap 5.9d · longest 68d · 26% of past waits were shorter than this one
$ The OpenAI Codex team periodically wipes usage for all paying users, announced only on @thsottiaux's X account. Every one of them is logged here.
// ACTIVITY · 52 WEEKS UTC
regular banked policy
// LOG · 73 entries
newest firstDay 21/ Wonder what happens the remaining seven days. But today is a reset.
https://x.com/thsottiaux/status/2114588174249071102It is all propagated. Enjoy some mad codexing tomorrow. Fun week ahead too. — confirms the reset announced earlier (x.com/thsottiaux/status/2114200152701575656): Day 20/ Rest and reset By midnight PST
https://x.com/thsottiaux/status/2114237083652882932Day 18/ Decisions API is GA, OpenResponses joined the Linux Foundation, we shipped some performance improvements and Uber is going all-in on codex & work. But yes… that’s a reset! Landing tomorrow before 10am PST together with a small ship.
https://x.com/thsottiaux/status/2113492187770175603Day 17/ Bet you thought we forgot, but no, it’s reset day today. Also good time to remind you to get Daybreak Blue access, it only takes 15 mins and you don’t want to have to do it in the middle of an incident. Lands tomorrow before 10am PST.
https://x.com/thsottiaux/status/2113162277688070445Day 16/ You can now cache for 12h in the API and for logged in users across Codex & Work we keep cache forever and you don’t have any penalty for coming back to an old conversation or resuming after a few hours and are now guaranteed to hit the cache.
https://x.com/thsottiaux/status/2112712984644784534There were reports of usage draining faster than anticipated for some users over the weekend. We've identified approximately 40,000 accounts that had an inconsistency on which usage limits where the actual usage that was applied did not correspond to their actual selected plan. We are reaching out individually to all these users and have corrected the issue. If this affected you, we'll add credits to your account. Sorry we got this wrong.
https://x.com/thsottiaux/status/2112592264686149749Day 15/ Nevermind, we're still cooking... and reset. Enjoy.
https://x.com/thsottiaux/status/2112432076767289678// NOTIFY ME
Email me when Codex resets
One email per official reset. No digest, no marketing, one-click unsubscribe. Prefer X? Reset alerts are also posted at @XIAOMIAO_26 · RSS
// PLANS
| Plan | Price | 5h gate |
|---|---|---|
| Free / Go $0 Active Enough to try it out | ||
| Plus $20/mo Active Entry allowance; heavy debugging or large file reads hit the cap quickly | ||
| Pro $100/mo Disabled About 5× Plus | ||
| Pro $200/mo Disabled About 20× Plus, sized for a full working day | ||
| Business / Enterprise per seat Active Each seat defaults to roughly Plus level | ||
| Pro $500/mo Disabled Highest tier with exclusive Ultrafast mode (~8× faster generation in Codex/Work) | ||
Since 2026-04 usage is metered in tokens rather than message count, so no fixed message cap is published.
// WHERE TO CHECK
Codex CLI
Type /status in a session, or /usage to see banked resets and when they expire
ChatGPT desktop
Sidebar → avatar → Usage shows both windows and lets you redeem a banked reset
Web
Settings → Billing shows usage and billing history
// ERRORS
// FAQ
01 When does the Codex 5-hour window reset?
The window opens the moment you send your first message and resets 5 hours later. It is not pinned to the clock and is not synced across users — every account has a different start point depending on when you began.
02 Does the weekly allowance reset on my billing date?
No. The weekly window starts from your first request after the previous reset and resets automatically 7 days later, independent of your billing date. Many people assume renewing clears it — it does not. This is a recurring point of confusion in official issues.
03 Is the 5-hour limit the same on Plus and Pro?
No. Plus is constrained by both the 5-hour window and the weekly allowance; Pro ($100 / $200) had the 5-hour gate removed in August 2026, leaving only the weekly allowance. Business / Enterprise defaults to roughly Plus-level per seat.
04 What is a banked reset?
A saved reset credit issued by OpenAI. It does not refill immediately — you redeem it yourself in the Codex app. Redeeming resets both the 5-hour and the weekly window. Credits expire 30 days after they are granted.
05 Is a paid instant reset extra allowance?
No. Plus / Pro can buy an instant reset from Usage settings, but it pulls your next period's allowance forward rather than adding a separate entitlement on top.
06 How do I check what I have left?
Type /status in the Codex CLI to see remaining usage; /usage shows earned reset credits and their expiry. On desktop, click your avatar in the sidebar → Usage to see both windows.
07 Did Codex reset today?
Check the "last reset" time at the top of this page — it updates whenever OpenAI announces one. If it says hours ago, yes. For confirmation inside the app, /status in the Codex CLI shows whether your windows are back at 100%. Resets are announced by @thsottiaux on X, usually within an hour of going live.
08 Can I predict when the next Codex reset will happen?
Not reliably — there is no schedule. But the history above shows what triggers them: user milestones (7M, 10M, 20M…), outages and incidents (a reset almost always follows), model launches, and fixes to usage metering bugs. The average gap in our records is about a week, but they cluster around news. Anyone selling a "probability" is guessing from the same public data.
09 Why is my Codex usage depleting faster than expected?
In August 2026 OpenAI found and fixed eight separate causes: compaction keeping old images, memory workers that never stopped, /goal runs overshooting their stop condition, automations firing too often, subagents escalating to bigger models, computer history re-summarising, background task summaries, and double-encoded MCP results. If you still see it, /status shows where usage goes. Reports on X or in the GitHub repo do get investigated — several past resets were compensation for exactly this.
10 What is the difference between the 5-hour limit and the weekly limit?
They are two independent gates. The 5-hour window caps how much you can do in a burst and rolls from your first message; the weekly allowance caps the total across 7 days. On Plus you can hit either one first. Pro plans dropped the 5-hour gate in August 2026, so only the weekly cap applies there. A reset from OpenAI clears both.
11 Why did my Codex 5-hour limit disappear?
Two possible reasons. If you are on Pro, it was removed permanently in August 2026. If you are on Plus, OpenAI temporarily lifted it in July 2026 so people could push GPT-5.6 Sol — it came back later. When the gate is off, /status only shows the weekly bar.
12 Who announces Codex resets?
Tibo (@thsottiaux), who leads Codex at OpenAI, posts almost every one on X — usually a casual one-liner, sometimes with a milestone or an apology attached. There is no official status page or email for resets. This site records each announcement with the original link.
13 What was the "2x" Codex limit promotion?
OpenAI has run 2× allowance promotions more than once: over the 2025 holidays (25 December to 1 January) and again in early 2026. In March 2026 they found the 2× multiplier had not been applied to about 9% of Plus and Pro users and issued a reset as compensation. When a promotion ends your bar looks like it "dropped" — it went back to standard.
14 What time does Codex reset?
There is no clock time for the normal windows: the 5-hour window and the weekly window both roll from your own first request, so /status is the only place to see your exact time. Announced fleet-wide resets are different — Tibo usually says when they land, and it is almost always "end of day" or "by midnight" Pacific time (07:00 UTC during daylight saving, 08:00 in winter), occasionally "within the hour". The last-reset time at the top of this page is when the reset was confirmed live, shown in your local time zone.
15 How many banked Codex resets can I have, and when do they expire?
They stack. In September 2026 OpenAI issued one banked reset for every day a paying user was without Astra access, so some accounts held several at once — the profile menu shows the count, e.g. "3 resets available". Each one expires 30 days after it was issued and unused ones simply lapse; OpenAI does not compensate for a reset you let expire. Redeeming one refreshes both the 5-hour and the weekly window and moves your weekly reset date.
16 My Codex reset disappeared — where did it go?
Three usual explanations. You redeemed it (/usage shows the history), it hit its 30-day expiry, or it was one of the banked resets that failed to apply on 9 September 2026 — OpenAI re-issued those and sent an apology email. If none of those fits, contact OpenAI Support with the count you saw and when; they will not grant a fresh reset on request, but they do investigate ones that were not applied.
17 Does a Codex reset also reset ChatGPT Work?
Yes. Codex, ChatGPT Work, ChatGPT for Excel and Workspace Agents draw from one shared allowance on your plan, so a fleet-wide reset — or a banked one you redeem — clears them together. That is why Tibo's announcements often say "Codex and ChatGPT Work users". Regular ChatGPT chat is metered separately and is not affected.
18 What time does Codex reset in Beijing, Tokyo or Seoul time?
Only fleet-wide resets have a clock time; your own windows roll from your first request. When Tibo says "by midnight Pacific", that is 07:00 UTC in summer (08:00 in winter) — 15:00 Beijing / 16:00 Tokyo and Seoul in summer, an hour later in winter. "End of day Pacific" lands around the same time. The last-reset time at the top of this page is already in your own time zone, so there is nothing to convert.
19 Does reasoning effort or a long context make me hit the Codex limit faster?
Yes. Since April 2026 Codex meters tokens, so anything that produces more output (high reasoning effort, long agent runs, many tool calls) or re-sends more input (a long conversation, large files in context) drains both windows faster. Cached input is cheaper but not free. /status shows where usage went; starting a fresh session for a new task is the cheapest fix.