OpenAI Resets ChatGPT Work And Codex Limits For Paid Users
OpenAI reset usage for paid ChatGPT Work and Codex users on August 8 and August 29, 2026, and for all paid subscriptions on September 7. OpenAI's help center says these resets are one-time, apply automatically, and are not guaranteed to recur. Check Settings, the Codex usage dashboard, or /status in the CLI for your own reset time.
OpenAI has reset usage for paid ChatGPT Work and Codex users three times since August 2026. The most recent was a global reset for all paid subscriptions on September 7, confirmed by OpenAI executive Tibo a few hours later: "All reset for everyone. Enjoy the week with Astra." (Tibo on X)
Each reset was one-time. OpenAI's help center says an automatic reset is applied directly to the covered limits, is not saved for later, and that "Future resets are not guaranteed." To see where your account stands, open Settings or the Codex usage dashboard, or enter /status in a Codex CLI session. (OpenAI Help Center)
Reset Timeline
Three resets and one limit change were announced between August 8 and September 8, 2026. All four came from the X account of OpenAI executive Tibo, not from an OpenAI blog post or release note.
| Date (UTC) | What was announced | Who it names |
|---|---|---|
| August 8 | "I have reset usage limits for all paid users of ChatGPT Work and Codex." | Paid ChatGPT Work and Codex users |
| August 25 | The five-hour limit returns "Tomorrow" for Plus accounts | Plus; Pro $100 and Pro $200 stay without it |
| August 29 | "We are reseting usage for all paid users of Codex and ChatGPT Work", with eight usage fixes | Paid Codex and ChatGPT Work users |
| September 7 | "We will do a global reset of the usage for all paid subscriptions" | All paid subscriptions |
| September 8 (04:05) | "All reset for everyone. Enjoy the week with Astra." | Everyone covered by the September 7 post |
Key Takeaways
- OpenAI reset paid ChatGPT Work and Codex usage on August 8 and August 29, 2026, and all paid subscriptions on September 7.
- The resets are one-time. OpenAI's help center says future resets are not guaranteed and Support does not reset limits on request.
- The August 29 post listed eight fixed usage bugs and said Codex usage should go between 10% and 50% further than before, depending on how you use it.
- The five-hour limit came back for Plus accounts in late August. Tibo said Pro $100 and Pro $200 stay without it for the upcoming months.
- Codex, ChatGPT Work, ChatGPT for Excel, and Workspace Agents share one allowance and credit pool.
- None of the posts mentions the OpenAI API.
- Check Settings, the Codex usage dashboard, or
/statusin the CLI for your own limits and reset time.
What To Do Now
Your account's usage page is the only authoritative answer to "was I reset, and when do I reset next?". The posts do not give per-account results.
| Your situation | What to do |
|---|---|
| You want to know if a reset reached you | Open Settings or the usage dashboard and read the allowance and reset time shown |
| You are in a Codex CLI session | Enter /status |
| You hit a limit mid-task | Codex can finish the active turn, subject to fair-use limits; then check the limit notice |
| You need more usage now | The limit notice lists your options: add credits, use an available reset, upgrade, or wait. Eligible Plus and Pro users can buy credits without changing plans |
| Your numbers look wrong, or access did not return after the stated reset time | Contact OpenAI Support with the limit or balance in question, the reset time shown, a screenshot, the client and model, and your time zone |
| You want Support to reset your limit | They will not: "OpenAI Support does not reset ChatGPT or Codex usage limits" |
Every row comes from OpenAI's help article. (OpenAI Help Center)
The August 8 Reset
The first reset was announced alongside the GPT-5.6 Sol rollout. Tibo wrote: "I have reset usage limits for all paid users of ChatGPT Work and Codex." The same post says GPT-5.6 Sol "can be used pretty much anywhere, including in the CC harness", a reference to running the model inside Claude Code; the post quote-tweets an Anthropic engineer saying the company does not ban people for using harnesses with other models. It states no new allowance, expiration time, or change to plan terms. (Tibo on X)
The August 29 Reset And The Usage Fixes
The second reset came with an explanation of why Codex usage had been draining faster than expected. Tibo wrote that the team had been "going through thousands of reports and shipping fixes" and that, "Depending on how you use Codex, you should see your usage go between 10% and 50% further than before." (Tibo on X)
The post lists eight problems, each marked fixed or disabled:
| Area | What the post says was wrong |
|---|---|
| Compaction | Old images were kept during compaction, sometimes making the context large enough to trigger compaction again; usage dropped around 10% for heavy image users after the fix |
| Memory | Background memory workers could inherit Stop hooks and keep running; fewer than 1% of users affected |
| Goals | A set /goal could keep going past its stop condition or keep retrying broken tools; examples consumed 15% to 70% of a weekly allowance |
| Automations | Some custom schedules could run more often than configured |
| Subagents | Smaller models sometimes picked more capable helpers unasked, and sub-agents could be requested in /fast when the orchestrator was not |
| Computer History | Overlapping activity was summarized repeatedly; in some cases up to one fifth of weekly usage |
| Rolling task summaries | Ordinary turns triggered extra background requests, about 1% of token usage; now disabled |
| MCP | Some tool results could be encoded twice, and cut-off tool instructions were fetched again |
The post adds that OpenAI is working on showing in the app where your usage goes. If you use /goal, see Codex CLI 0.128 and persisted goals; for the later Astra efficiency claim, see GPT-6 Astra long-tail usage improvements.
The September 7 Global Reset
The third reset was the broadest. Tibo wrote: "We will do a global reset of the usage for all paid subscriptions so that you can keep enjoying Astra", and added "Lands around 6pm PST today." The confirmation followed at 04:05 UTC on September 8. (Tibo on X; follow-up)
The post names "all paid subscriptions" rather than ChatGPT Work and Codex specifically. It does not say whether limits outside the shared agentic pool were included, so read your own usage page rather than assuming.
The Five-Hour Limit For Plus
The five-hour limit is a plan change, not a reset. In a post timestamped August 25, 2026 at 01:16 UTC, Tibo wrote: "Tomorrow we will bring back the 5h limit for Plus accounts across ChatGPT Work and Codex." Depending on the time zone he meant, that is August 25 or August 26. (Tibo on X)
He gave two reasons: the limit lets OpenAI smooth the load on its compute while keeping the plan generous in weekly usage, and Plus users "accidentally eat through their whole weeks usage". He also wrote: "We are for the upcoming months keeping the 5h limit not enabled for Pro $100 and Pro $200 subscriptions."
Two OpenAI-owned pages are consistent with a five-hour window on Plus. The Codex CLI 0.153.0 release notes (September 3, 2026) say "Plus and Team users receive an earlier warning when less than half of their allowance remains in an approximately five-hour usage window." The help center says a banked reset "refreshes your 5-hour and weekly Codex usage windows and changes your weekly reset date." (Codex CLI 0.153.0; OpenAI Help Center)
How Much Usage A Plan Includes
The reset posts publish no quota, but OpenAI's Codex pricing page does publish estimates. As of September 19, 2026 it gives these ranges of local messages per five-hour period, and says they "are not fixed message limits" and that "Weekly limits may also apply." (Codex pricing)
| Model | Plus | Pro 5x | Pro 20x |
|---|---|---|---|
| GPT-6 Astra | 5-45 | 25-225 | 100-900 |
| GPT-5.6 Sol | 10-100 | 50-500 | 200-2,000 |
| GPT-5.6 Terra | 25-200 | 125-1,000 | 500-4,000 |
| GPT-5.6 Luna | 250-2,000 | 1,250-10,000 | 5,000-40,000 |
Standard Business seats carry the Plus ranges. The page expresses every plan's estimate per five-hour period, including Pro; it does not say which plans enforce a five-hour window, so the Pro statement above stands on Tibo's post alone.
How ChatGPT Work And Codex Usage Relate
Work and Codex draw on one pool, so activity in either reduces the same allowance. As of September 2026 the help center says: "Codex, ChatGPT Work, ChatGPT for Excel, and Workspace Agents use a shared allowance and credit pool when those features are available on your plan." Usage depends on the model, where the task runs, task complexity, context, reasoning, speed, and tools, and a long-running task can use substantially more than a short request. (OpenAI Help Center)
Task work started through ChatGPT Voice in Work or Codex on desktop uses the same allowance. Regular Chat Voice has separate limits and does not consume Codex usage. ChatGPT file upload, image, and voice caps are separate from Codex limits.
Automatic Resets Versus Banked Resets
OpenAI documents two kinds of reset, and the three announcements above were the automatic kind.
| Automatic one-time reset | Banked reset | |
|---|---|---|
| How you get it | OpenAI applies it to the limits an announcement covers | Some promotions, such as Plus and Pro referrals |
| When it takes effect | Immediately; it "is not saved for later" | When you choose, from the usage summary in the profile menu |
| What it refreshes | The limits covered by the announcement | Your 5-hour and weekly Codex windows; it also changes your weekly reset date |
| Expiry | None to track; usage after it counts toward the refreshed allowance | Set by the offer; check your account for redemption and expiration details |
OpenAI says it does not provide an additional reset, credit, refund, or other compensation because you were not eligible for a one-time reset, did not notice one, or did not use a banked reset before it expired. (OpenAI Help Center)
What The Resets Do Not Establish
| Not established | Why |
|---|---|
| A recurring reset schedule | The help center says future resets are not guaranteed |
| A permanent increase to included usage | The August 29 post describes bug fixes that make usage go further, not a larger plan allowance |
| API credits or API rate-limit changes | No post mentions the API |
| That every account received the same result | The help center says eligibility and the limits affected may vary by offer |
Bottom Line
OpenAI reset paid ChatGPT Work and Codex usage on August 8 and August 29, 2026, and all paid subscriptions on September 7. The August 29 reset came with eight usage fixes that Tibo said should make Codex usage go 10% to 50% further. The five-hour limit is back on Plus and, per the same account, not on Pro $100 or Pro $200 for now. Resets are one-time and not guaranteed, so plan around the usage dashboard, not around the next announcement.
Corrections and Updates
- September 19, 2026: This page said the August 25 post named only Plus accounts and warned against inferring anything for other plans. The post also says the five-hour limit stays off for Pro $100 and Pro $200 subscriptions for the upcoming months. The page also gave August 26 as the return date; the post says "Tomorrow" and is timestamped August 25, 01:16 UTC. It said the August 29 post gave no numbers; the post says usage should go 10% to 50% further and lists eight fixes. It said the September 7 post gave no reset mechanics; the post says "Lands around 6pm PST today."
Sources
- Tibo on X, "I have reset usage limits for all paid users of ChatGPT Work and Codex" (August 8, 2026): https://x.com/thsottiaux/status/2086188036493344823
- Tibo on X, "Tomorrow we will bring back the 5h limit for Plus accounts across ChatGPT Work and Codex" (August 25, 2026): https://x.com/thsottiaux/status/2092058556707344708
- Tibo on X, "We are reseting usage for all paid users of Codex and ChatGPT Work" (August 29, 2026): https://x.com/thsottiaux/status/2093801758665715784
- Tibo on X, "We will do a global reset of the usage for all paid subscriptions" (September 7, 2026): https://x.com/thsottiaux/status/2097043464538264003
- Tibo on X, "All reset for everyone. Enjoy the week with Astra." (September 8, 2026): https://x.com/thsottiaux/status/2097174560412246215
- OpenAI Help Center, "Using Codex with your ChatGPT plan": https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan
- OpenAI Developers, Codex pricing: https://developers.openai.com/codex/pricing
- OpenAI, Codex CLI 0.153.0 release notes (September 3, 2026): https://github.com/openai/codex/releases/tag/rust-v0.153.0
Read next
Related coverage- OpenAI
OpenAI Launches ChatGPT Work: A Codex-Sibling Agent for Long-Running Deliverables
OpenAI introduced ChatGPT Work, a new agent in ChatGPT powered by Codex and GPT-5.6 that takes action across apps and files and stays on a project for hours to turn a goal into finished work. Work handles research and deliverables; Codex stays the dedicated software-development agent.
- OpenAI
OpenAI Brings Codex Security Review to GitHub Pull Requests
Codex Security Review is OpenAI’s research-preview workflow for deeper security analysis of GitHub pull requests. It uses the diff, repository context, and optional threat-model guidance, then reports actionable findings in the pull request and a fuller report in Codex. Enterprise, Business, Edu, and Pro users can configure it; Plus is excluded.
- Codex
Codex Permission Profiles: Least-Privilege Controls for Local Agent Work
OpenAI shipped Codex permission profiles in beta -- reusable, inheritable policies that replace the coarse sandbox_mode/sandbox_workspace_write combo. A profile binds OS-enforced filesystem read/write/deny rules (down to **/*.env) to per-domain network and Unix-socket rules. Enterprise admins get fail-closed allowlists via requirements.toml. Profiles govern local sandboxed command execution only, not MCP servers, app connectors, browser, or cloud.
Frequently Asked Questions
What did OpenAI reset?
OpenAI executive Tibo posted three resets in 2026: August 8 and August 29 for all paid users of ChatGPT Work and Codex, and September 7 for all paid subscriptions, confirmed a few hours later with "All reset for everyone." Each was a one-time reset of current usage, not a change to what a plan includes.
What did OpenAI say about the August 29 reset?
The August 29 post paired the reset with a list of eight fixed usage bugs, in compaction, memory workers, /goal, automations, subagents, Computer History, rolling task summaries, and MCP. Tibo wrote that, depending on how you use Codex, usage should go between 10% and 50% further than before. It did not publish a new plan quota.
What five-hour limit change did OpenAI announce?
In a post timestamped August 25, 2026 (01:16 UTC), Tibo said the five-hour limit would return the next day for Plus accounts across ChatGPT Work and Codex, to smooth compute load and stop casual users from spending a whole week of usage by accident. He said Pro $100 and Pro $200 subscriptions would stay without it for the upcoming months.
Does the reset apply to the OpenAI API?
No post mentions API accounts. The August posts name paid ChatGPT Work and Codex users, and the September 7 post names paid subscriptions. OpenAI's Codex pricing page lists API key usage separately as usage-based, so do not read the resets as an API credit, billing, or rate-limit change.
Do ChatGPT Work and Codex share usage?
Yes. As of September 2026, OpenAI's help center says Codex, ChatGPT Work, ChatGPT for Excel, and Workspace Agents use a shared allowance and credit pool when those features are on your plan. Usage depends on the model, where the task runs, task complexity, context, reasoning, speed, and tools.
How can I check my remaining Codex usage?
Open Settings or the Codex usage dashboard at chatgpt.com/codex/settings/usage and check which allowance is exhausted, your credit balance, and the reset time shown. In an active Codex CLI session, enter /status. The limit notice lists your options: add credits, use an available reset, upgrade, or wait.
Will OpenAI reset usage again, and can Support reset my limits?
OpenAI's help center says one-time resets are occasional, eligibility varies by offer, and "Future resets are not guaranteed." It also says OpenAI Support does not reset ChatGPT or Codex usage limits, and that no extra reset, credit, or refund is given because you missed or did not notice one.