Codex usage limit reached: what to do

What does “You’ve hit your usage limit” mean in Codex?

Your ChatGPT plan has a usage allowance for Codex, and you used it up. When a Codex usage limit is reached, the CLI prints “You’ve hit your usage limit” and usually a time to try again, in your local time. The wording follows your plan: on Plus it points to Pro and to buying more usage, on Pro to buying more usage, on Free and Go to Plus, and on Business to your admin. That is how the open-source CLI builds the message, per its error source.

Two kinds of limit can apply. OpenAI’s pricing page gives Plus limits per five-hour period and adds that weekly limits may also apply. It says local messages and cloud chats share one allowance, so a cloud task you started this morning draws on the same budget as the session in front of you.

Some limits are tied to a named bucket, such as one model. The error source has a separate message for that case: “You’ve hit your usage limit for” that name, then a prompt to switch to another model. If you hit the limit in the middle of a turn, OpenAI says the agent can finish that turn, subject to fair use limits, per the pricing FAQ.

Where can you check your Codex usage and reset time?

In an active CLI session, /status shows your remaining limits, and the usage dashboard shows current limits and reset times, according to OpenAI’s pricing page. The same page suggests checking the dashboard every week or two to learn your pace.

The CLI has more. /usage opens a usage menu where you can show token activity or redeem an earned reset when one is available, and /usage daily, /usage weekly or /usage cumulative open token activity directly. /statusline lets you keep your rate limits in the footer, so you see the number before you hit the wall. The CLI command docs list all three.

The pricing page does not publish a fixed reset schedule. Take the reset time from the message or the dashboard, not from a guess.

What can you do right now?

Pick a path:

Wait. The message gives a time to try again, and the dashboard shows reset times.

Buy more usage. OpenAI’s pricing FAQ says Plus and Pro users who reach their limit can buy more usage and keep working without changing plans. Business, Edu and Enterprise plans with flexible pricing can buy more usage for the workspace.

Use an API key. The same FAQ says all users can run extra local chats with an API key, billed at standard API rates. printenv OPENAI_API_KEY | codex login --with-api-key signs the CLI in, per the authentication docs. Codex then uses API pricing instead of your plan’s included usage, and the pricing page lists no cloud features, such as GitHub code review, for the API key option.

Move up a plan. As of October 2026, OpenAI’s pricing page lists Pro from $100 a month and says Pro plans currently have no five-hour limit.

Change the model or the speed. OpenAI suggests asking whether a smaller model or a tighter task would still give a useful result. /model switches models. On models that support it, /fast toggles Fast mode, which uses included limits at 2.5 times the standard rate, so turn it off when usage is short. See the speed docs.

Switch tools for the rest of the day. More on that below.

What are Codex usage limits by plan?

OpenAI’s pricing page lists Codex in Free ($0), Go ($8 a month), Plus ($20 a month), Pro (from $100 a month), Business, and Enterprise or Edu plans, plus an API key option billed at API rates. It describes Plus as a plan to “power a few focused coding sessions each week.”

Plus. OpenAI estimates local messages per five-hour period by model: 5 to 45 on GPT-6 Astra, 15 to 150 on GPT-6 Sol, 15 to 160 on GPT-6.1 Sol and 350 to 3,000 on GPT-6 Luna. It calls these estimates, not fixed message limits, because model, context, reasoning and tool use change what a message costs.

Pro. The page says Pro plans currently have no five-hour limit, and lists plans at $100, $200 and $500 a month.

Free and Go. The page gives no message numbers for either.

Business and Enterprise. Standard Business uses the same estimates as Plus. Enterprise and Edu plans with flexible pricing have no fixed rate limits, and usage grows with what the workspace buys.

Figures and model names move fast. The same page says GPT-5.5 retired from Codex with ChatGPT sign-in on October 14, 2026, so check it before you plan around a number.

Why did Codex run out so fast?

Usually because of what each message carried, not how many you sent. OpenAI says tasks that look similar can use different amounts of your allowance, because model choice, context, reasoning, tool use, retrieval and caching all affect usage. Prompt length alone is not a reliable estimate.

A few multipliers are documented. Cloud tasks may use more of your allowance than local messages. Fast mode uses included limits at 2.5 times the standard rate on supported models. Image generation uses included limits about 3 to 5 times faster on average. All three are on the pricing page.

OpenAI’s own tips for making limits last are plain. Send precise instructions. Give the agent only the files it needs. Keep AGENTS.md small by nesting it in your repository. Turn off MCP servers you are not using, because each one adds context to your messages. See its list. Our guide to controlling AI coding costs covers the same habits for any agent.

How do Codex usage limits compare with Claude Code’s?

Not number for number, because the vendors count differently. OpenAI publishes message estimates for Plus. Anthropic’s pricing page says there is no fixed message count and describes plans in multiples: Pro gives at least 5 times Free’s usage per five-hour session, and Max gives 5 or 20 times Pro’s.

The shape is similar. On the entry plans both can use a five-hour window with weekly limits on top, and each shares one allowance across surfaces: OpenAI across local messages and cloud chats, Anthropic across Claude and Claude Code. ChatGPT Plus is $20 a month on OpenAI’s page, Claude Pro is $20 billed monthly or $17 billed annually on Anthropic’s, and both vendors’ next tiers start at $100.

So compare on your own work. Run the same task in each tool, then read /status in Codex and /usage in Claude Code. The Claude Code vs Codex vs Gemini CLI comparison covers the other differences. If it is Claude Code that stopped you, read what to do when Claude Code hits its limit, and Claude Max vs Pro if you are weighing an upgrade.

How do you keep working with another agent?

The limit belongs to your OpenAI account, not to your project. Claude Code and Gemini CLI have their own allowances, so the work can continue there, though Gemini CLI’s changed in June 2026: read Gemini CLI limits before you count on it.

A handoff is only as good as the brief. Run git status and commit or stash what works. Then write five lines: the goal, what is done, what is left, the failing check and the files that matter. Paste them into the next agent with the test command, and ask it to run the check before it edits anything. Running Claude Code and Codex together covers a longer setup.

Hand the work to another agent

SwarmPane’s Limit handoff is built for this moment. When a usage limit hits, you can switch to your other account, or from Codex to Claude Code and back, with a brief of the work that you edit first, or resume after the reset, even if your computer slept. Usage shows one row per CLI from its own reports, including Codex’s live limits when Codex shares them, and Accounts saves several Claude Code and Codex sign-ins so you can pick one for each pane. Use accounts you are entitled to use, and follow each provider’s terms.

SwarmPane runs the agent CLIs and accounts you already have, so the providers bill their own usage separately. Start with a 7-day trial for $1 and cancel before it ends if it is not for you.