Codex usage limits explained (2026)

Codex limits come from your ChatGPT plan: Plus ($20 a month) has a five-hour window and weekly limits, and OpenAI's pricing page says Pro plans currently have no five-hour limit. The numbers depend on the model and the task, so run /status in Codex to see yours.

Contents

Codex usage limits come from your ChatGPT plan. Plus ($20 a month) has a five-hour window and weekly limits; for Pro ($100, $200 or $500), OpenAI's pricing page says there is currently no five-hour limit. How many messages fit depends on the model and the task, and the numbers moved this year, so run /status in Codex and check OpenAI's page.

The shape: two windows#

Codex meters you on two clocks at once.

  • Five hours: a rolling window for a burst of work. When it's spent, Codex waits for the reset /status shows.
  • One week: the bigger ceiling. OpenAI's line is "Weekly limits may also apply", and for heavy users the week is the one that stops them.

Local messages (the CLI and the IDE extension) and cloud tasks draw on the same two. OpenAI says cloud tasks "may use more of your allowance than local messages".

How many messages is that?#

OpenAI gives a range per model, not a fixed number. The table is its estimate of local messages per five-hour period on Plus and on Standard Business seats, as the pricing page listed them on 22 October 2026.

Model Plus Standard Business
GPT-6 Astra 5–45 5–45
GPT-6.1 Sol 15–160 15–160
GPT-6 Sol 15–150 15–150
GPT-6 Luna 350–3,000 350–3,000

OpenAI adds that "these estimates are not fixed message limits". An agent turn that reads half your repo counts for far more than a one-line question, which is why GPT-6 Astra's range runs from 5 all the way to 45.

Why it drains faster than you expect#

The count, in OpenAI's words, "depends on the model used, size and complexity of your tasks, and whether you run them locally or in the cloud". Four things set how soon you reach the limit.

  • The model. GPT-6 Astra fits 5 to 45 local messages in a window, GPT-6 Luna 350 to 3,000.
  • Cloud tasks, which draw more from the shared allowance than local messages do.
  • Context. Every turn carries the conversation so far, so a message late in a long session counts for more than the same message in a fresh one.
  • Parallel runs. Three Codex sessions on one account drain one allowance about three times as fast.

Plans at a glance#

From OpenAI's pricing page, read on 22 October 2026, in US dollars:

Plan Price What OpenAI says about Codex
Free $0 "Explore Codex capabilities on quick coding tasks"
Go $8 a month "Use Codex for lightweight coding tasks"
Plus $20 a month "Power a few focused coding sessions each week". Five-hour window and weekly limits
Pro $100, $200 or $500 a month "Pro plans currently have no five-hour limit". Weekly limits may apply
Business $20 per user a month, billed yearly, 2+ users Same five-hour estimates as Plus for Standard seats
Enterprise and Edu Contact sales No fixed rate limits on flexible pricing; usage scales with credits

Plus and Pro users who hit a limit can buy credits from OpenAI and keep going.

How to check your own limits#

Type /status in a running Codex session for both windows and when each resets. On the web, OpenAI's pricing page links a usage dashboard with the same limits and reset times. ChatGPT's usage data can also carry a separate Code review limit, which Shuriken shows as a third bar.

Look before a long job, not after: a refactor started at 10% of the week can stop halfway through the migration.

What changed in 2026#

  • 12 July 2026: the five-hour limit was lifted for a while. OpenAI Support on the developer forum, 28 July: "As announced on July 12, 2026, the 5-hour Codex usage limit has been temporarily lifted for Plus, Pro, and Business users." Weekly limits stayed.
  • 29 July: a post in the same forum thread said the limit would return the next day. On 1 August the same poster wrote that it had been postponed.
  • 25 August 2026: the five-hour limit came back for Plus. Notebookcheck reported that the $100 and $200 Pro tiers stayed exempt "over the coming months", citing a post by Thibault Sottiaux, whom it describes as OpenAI's engineering lead for Codex and ChatGPT.
  • 22 October 2026, when we last checked: OpenAI's pricing page says "Pro plans currently have no five-hour limit."

Plus went 44 days without the five-hour limit. Anything you build around Codex limits should read the real ones (/status, or the dashboard) instead of a figure from a blog post, this one included.

Plus or Pro for Codex?#

Plus is sized, in OpenAI's words, to "power a few focused coding sessions each week". A full day in Codex meets the five-hour wall, and parallel agents meet it sooner because they all draw from the same window.

Pro removes the five-hour wall for now and raises the ceiling, which leaves the weekly limit as the number to watch.

Neither fits a bill that is mostly one or two big automated runs. An API key bills per token and has no plan windows.

Against Claude Code, Claude Code vs Codex compares the two agents and Claude Max vs ChatGPT Pro for coding the two top plans.

When the weekly limit is used up#

Your options, cheapest first:

  1. Wait. /status shows when the week resets.
  2. Use a banked reset if your account has one. The Codex limit reset tracker shows whether OpenAI is handing out a free one right now.
  3. Buy credits, if your plan offers them.
  4. Switch Codex to an API key and pay per token until the plan resets.
  5. Hand the task to another agent you already pay for, such as Claude Code.
  6. Use another ChatGPT account you own (next section).

Several ChatGPT accounts#

Two accounts, say a work Business seat and a personal Plus, or a Plus and a Pro, means two sets of windows.

Codex itself holds one sign-in per home folder. OpenAI's auth docs put the credentials in auth.json under CODEX_HOME (~/.codex by default) or in your system's credential store, so a second account by hand means a logout or a second home.

  1. codex logout, then codex login as the other account, every time you switch.
  2. CODEX_HOME=~/.codex-work codex, with one folder per account. Each folder has its own config.toml, so MCP servers and settings need copying or linking; switching Codex CLI accounts without logging out links config.toml and the sessions folder from ~/.codex into each extra home.

Only use accounts you own and are entitled to use. Providers' terms may not allow sharing a subscription sign-in, and it can put your account at risk.

In Shuriken, each ChatGPT sign-in is a source (Codex with several ChatGPT accounts). Sources shows every account's Session and Weekly bars side by side, plus Code review where ChatGPT reports it. When the account in use reaches your switch point (99% by default, adjustable from 50% to 100%) or OpenAI reports a limit, your chats move to the account with the most room, on every device at once (source switching). A Codex chat moves when its current turn ends, and a "Switched to …" card in the chat says which account took over.

Download Shuriken and add your ChatGPT accounts under Sources.

Sources

  1. OpenAI: Codex pricing (accessed 22 October 2026)
  2. OpenAI Developer Community: Has the 5-hour usage session been removed from Codex CLI? (OpenAI Support reply, 28 July 2026) (accessed 22 October 2026)
  3. Notebookcheck: OpenAI restores 5-hour Codex and Work limits for ChatGPT Plus (25 August 2026) (accessed 22 October 2026)
  4. OpenAI: Codex authentication (accessed 22 October 2026)

Frequently asked questions

What's the difference between Plus and Pro for Codex?

On 22 October 2026, OpenAI's pricing page described Plus ($20 a month) as enough for a few focused coding sessions a week, with a five-hour window and weekly limits. It said Pro plans ($100, $200 or $500 a month) currently have no five-hour limit. Weekly limits may still apply.

Is the Codex 5-hour limit back?

For Plus, yes. OpenAI lifted it temporarily on 12 July 2026, and it returned for Plus on 25 August 2026, per press reports of the Codex lead's announcement. OpenAI's pricing page says Pro plans currently have no five-hour limit.

My Codex weekly limit is used up. What can I do?

Wait for the reset shown in /status, buy credits if your plan offers them, switch Codex to an API key, or continue on another account you own. Shuriken can move your chats to another of your ChatGPT accounts when one runs out.

How do I check my Codex usage?

Run /status inside a Codex CLI session, or open the usage dashboard OpenAI links from its pricing page. Both show your current windows and when they reset.

Do Codex cloud tasks use the same allowance?

Yes. OpenAI says local messages and cloud chats share your plan's usage allowance, and cloud tasks may use more of it than local messages.

Free tools