Codex usage limits: plans, resets and remaining usage
Quick answer: Codex usage limits depend on your plan and workload; check Settings → Usage or /status in the CLI for remaining allowance and reset times, then use the continuation options available to your account (OpenAI guidance).
A Codex usage limit is easier to diagnose when you separate three questions: which account is signed in, which allowance is exhausted, and when that allowance refreshes. This guide explains that check before you pay for more usage. It also separates current documentation from the plan changes shown in the recorded walkthrough.
How Codex usage limits work by ChatGPT plan
OpenAI publishes estimates, not a universal number of tasks you can finish. Its guidance says the model, task and settings affect consumption. Where five-hour and weekly windows both apply, you need remaining allowance in both. Pro currently has no five-hour limit, but still has an included allowance. See OpenAI’s usage-window explanation.
| Plan | What the official guidance establishes |
|---|---|
| Free and Go | The public pricing page describes limited trial access; it does not publish a guaranteed task count for either plan. Codex pricing. |
| Plus | Expanded access, with model-dependent local-message estimates. Treat those ranges as estimates rather than a fixed quota. Codex pricing; usage guidance. |
| Pro | Pro 100, Pro 200 and Pro 500 have increasing included usage; Pro 500 has the most. There is no verified universal weekly task count here. Pro tier details. |
| Business | Standard Business appears alongside Plus in the model-estimate table. Check the workspace’s actual seat and allowance before choosing it for more capacity. Codex pricing. |
| Enterprise and Edu | Flexible pricing scales with credits rather than fixed rate limits. Without flexible pricing, most features have Plus-equivalent per-seat limits. Codex pricing. |
Do not translate a plan name into “unlimited coding.” Compare your actual workload with the allowance shown in your account. A short correction and an open-ended investigation are different jobs; record what you asked for before comparing two sessions.
Local messages and cloud chats share the plan allowance, so moving a task to the cloud does not create a separate quota. OpenAI also says weekly limits may apply. Source: Codex pricing.
Where to see remaining usage
Open Settings or the Codex usage dashboard and check the exhausted allowance, available credit balance and displayed reset time. During an active CLI session, enter /status. These are OpenAI’s documented places to inspect the limit. Source: using Codex with your plan.
- Write down the exact limit message and the time you saw it.
- Check the account and workspace you intended to use.
- Read each allowance separately rather than relying on one percentage.
- Keep the displayed reset time with your task notes so you can plan the next attempt.
If the screen seems inconsistent, avoid guessing from a previous screenshot. For a support request, collect the plan, task reference, date, timezone, model, reasoning level and speed setting. OpenAI asks for these details when investigating unexpected usage. Source: usage troubleshooting.
What to do when a limit hits
First preserve your work: note completed edits, checks already run and the next unfinished step. Then choose a continuation path. OpenAI says an active turn may continue after a limit is reached, subject to fair-use limits; afterward, the usage page or banner shows the available choices. Source: reaching a limit during work.
Wait for the scheduled reset
Use the reset time shown for the exhausted allowance. On plans with both windows, the weekly and five-hour periods can end at different times; a refreshed weekly allowance does not necessarily mean the shorter window has refreshed. OpenAI’s reset guide explains this distinction.
Use the pause to inspect the diff, reproduce the error yourself or write a clearer next brief. That gives the next session a concrete starting point. Repeatedly sending the same broad request is unlikely to teach you which part of the task actually needs attention.
Review a plan change or eligible extra usage
The limit banner may offer an upgrade or additional credits. Availability depends on your account. On eligible Pro features, included usage is consumed before purchased credits; balances and purchases appear in Settings → Usage or Codex Usage & Billing. Source: Pro usage and billing. For credits and Pro tier details, read the existing plan and credit explainer.
A banked reset, if your account has one, is another option. Review its scope and expiry before redeeming it. Promotional resets are discretionary, and selected accounts may see paid instant resets; neither permanently raises the plan’s allowance. Source: reset types and eligibility.
Use separately billed API access for local work
OpenAI documents API-key sign-in for local work in the desktop app, CLI and IDE extension. This uses API billing through the OpenAI Platform account instead of included ChatGPT usage. Cloud Codex requires ChatGPT sign-in. Read the authentication guide before switching, and decide what you are willing to spend. An API key is a different billing path, not a subscription reset.
If you also use another coding assistant, the Claude Code vs Codex limit-handling section describes Mike’s fallback process. Keep a concise handover so the next attempt starts from the actual files and test results.
The recorded sign-in fix: April 2026
In my Codex usage limits: the recorded quota walkthrough, I hit my weekly limit and started a ChatGPT Business trial. The CLI still showed the old plan. Signing out and signing back in with the Business account was the fix that worked in that recording.
The plan details describe April 2026. The recording does not establish today’s trial availability, price or a guaranteed increase in weekly work. Its useful troubleshooting lesson is to verify which account the client is using after a plan change.
For a current authentication check, OpenAI documents codex login status; codex logout clears stored credentials, and codex login starts the ChatGPT sign-in flow. These commands concern authentication. Do not treat signing in again as a quota-reset method. Source: authentication and sign-out.
My later ChatGPT subscription-pause video describes the September recording, including my experience using up allowance and choosing medium reasoning for routine work. Those observations are one person’s experience, not a benchmark or a current plan comparison.
Practical ways to stretch your allowance
OpenAI recommends narrower prompts, relevant source files, appropriately sized output, smaller project instructions and fewer unnecessary MCP servers. Source: Codex pricing documentation. Put that advice into a brief you can check:
- Name the finish line. Ask for one broken behavior to be fixed, with a specific check showing it works.
- Point to relevant files. Include the failing command and error rather than asking for an unexplained review of everything.
- Separate required work from suggestions. Ask for optional improvements in the final notes instead of adding them to the implementation.
- Keep instructions useful. Remove stale project guidance and repeated background that does not help this task.
For routine work, OpenAI recommends a lower-cost model. Source: making usage last. Try that on a task with an obvious acceptance check, then inspect the result. There is no promised savings percentage in this guide.
Reasoning, context, tools and speed affect usage. Fast and Ultrafast consume included allowance faster than Standard for the same model. Source: different usage for similar tasks. Choose settings deliberately instead of assuming the most intensive option is necessary for every edit.
The Codex workflow guide gives you a small change to practice: brief it, inspect the diff and verify the result. Keep your own record of useful completed tasks, interruptions and rework. That is a better basis for your next subscription decision than copying someone else’s busiest day.
FAQ
Where can I check my remaining Codex usage?
Open Settings → Usage or the Codex usage dashboard. In an active CLI session, enter /status to see remaining limits.
Does Codex have a weekly limit?
Weekly limits may apply. Where both five-hour and weekly limits apply, you need allowance remaining in both windows.
Does starting a new chat reset the limit?
No. OpenAI says starting a new Codex chat does not reset usage limits.
Can I keep working after my included allowance runs out?
Check your limit banner for available options, such as waiting, upgrading, adding credits or applying an available reset. API-key access for local work is billed separately.
Will signing out and back in give me more usage?
The April recording shows signing in again fixing an old-plan display after a plan change. It does not establish signing out as a way to renew an exhausted allowance.
Sources
Official OpenAI sources, checked as of October 24, 2026. The developers documentation links may redirect to OpenAI’s current documentation host.