Limit almost used up, days before the reset: what to do?
Check two things first: what is left and when it resets (/usage in Claude Code, /status in Codex). Then cut back, fastest effect first: fewer subagents, lower reasoning effort, a smaller model, a new session for each new task. If that is not enough, the official ways out are to wait for the reset, upgrade your plan or buy extra usage; on Codex you can also use a banked reset.
Codex
/status shows what is left and when it resets. Cut subagents, lower effort, use a smaller model; then use a banked reset or buy usage.
First work out which limit you hit. The 5-hour limit comes back within hours; only the weekly limit needs rationing. Plus has a 5-hour limit; both Pro plans currently do not. Then cut back in this order: stop parallel subagents, lower the reasoning effort and pick a smaller model with /model, turn /fast off if it is on, and start a new session with /new when the task changes. What to do when that is not enough is in the section below.
What the official docs say
If you want to see your remaining limits during an active Codex CLI session, you can use /status.We are for the upcoming months keeping the 5h limit not enabled for Pro $100 and Pro $200 subscriptions.Because each subagent does its own model and tool work, subagent workflows consume more tokens than comparable single-agent runs.agents.enabled defaults to true. Set it to false to disable multi-agent tools.Higher effort can use more of your allowance and does not always produce a better result.Switch to a smaller model for routine tasks. Using GPT-5.6 Terra or GPT-5.6 Luna can extend your local-message usage limits, depending on the model you switch from.Claude
/usage shows what is left and when it resets. Stop subagents, lower effort, switch to Sonnet; then wait, upgrade or buy extra usage.
The Claude weekly reset time is fixed to your account: each reset is 7 days after the last one, whatever you use and whether or not a reset was announced. So the number of days you need to last is known. Cut back in this order: no subagent fanout, close agent teams when they finish, lower the effort with /effort, switch to Sonnet with /model, and run /rename then /clear when the task changes. Change the model and the effort at the start of a new session: changed in the middle of a long one, the next message re-reads the whole conversation without the cache.
What the official docs say
Weekly limits reset at a fixed time each week that is assigned to your account. Your reset day and time stay the same regardless of when you start using Claude or when your subscription begins, and you receive your full weekly allowance each cycle.A single burst of heavy activity, such as a large workflow fanout, can exhaust the weekly allowance before the session window resets.To prevent Claude from delegating to any subagent, deny the Agent tool itself with permissions.deny.Sonnet handles most coding tasks well and costs less than Opus. Reserve Opus for complex architectural decisions or multi-step reasoning. Use /model to switch models mid-session, or set a default in /config.Each model has its own cache. Switching with /model means the next request reads the entire conversation history with no cache hits, even though the content is identical.Use /clear to start fresh when switching to unrelated work. Stale context wastes tokens on every subsequent message.The session and weekly limits are shared across all models, so switching models doesn't restore access. The Opus and Sonnet limits each apply only to requests to that model family, so switching to a model outside the family with /model keeps you working.Cut back in this order
The steps are ordered by how fast they take effect. Do the first one before you touch the rest.
Subagents first. One large parallel fanout can use up a weekly allowance. In Codex, do not ask for parallel delegation, or write enabled = false under [agents] in ~/.codex/config.toml. In Claude Code, close agent teams when they finish; to block delegation completely, add Agent to permissions.deny in your settings.
Lower the reasoning effort one step: /effort in Claude Code, /model in Codex. Raise it again only for the hard problem.
Use a smaller model: Sonnet in Claude Code; in Codex, one of the smaller models the pricing page names. In Codex, turn /fast off if it is on.
New task, new session: /rename then /clear in Claude Code, /new in Codex. Change the model and the effort at this point, not in the middle of a long session.
If that is still not enough
Codex: if your account holds a banked reset, use it in Settings → Usage or with /usage in the CLI. Plus and Pro can also buy credits and keep working, or buy one immediate reset. After a banked or purchased reset, the next weekly reset is 7 days after the first message you send, so send one right away. OpenAI Support does not reset limits on request.
Claude: there are three official ways out: wait for the reset, upgrade the plan, or buy extra usage with /usage-credits. The session and weekly limits are shared across all models, so switching models after they run out does not bring access back. Claude has no banked resets.
On Claude, do not leave a session idle once you are on extra usage. The cache then lasts five minutes instead of an hour, and the first message after a longer break re-reads the whole conversation.
What the official docs say
ChatGPT Plus and Pro users who reach their usage limit can purchase additional credits to continue working without needing to upgrade their existing plan.Using a full banked reset refreshes your 5-hour and weekly Codex usage windows and changes your weekly reset date.A completed purchase immediately restores both 5-hour and weekly usage. It pulls your normal weekly allowance forward rather than adding a separate usage entitlement.OpenAI Support does not reset ChatGPT or Codex usage limits.If you hit your usage limit, you'll need to wait for it to reset, upgrade your plan, or purchase usage credits.Run /usage-credits to buy additional usage on Pro and Max, or to request it from your admin on Team and Enterprise.Cache misses: your first message after a break longer than the cache lifetime misses the cache and reprocesses your full context. The lifetime is an hour on a subscription and drops to five minutes once you're drawing on usage creditsCommon misconceptions
- The Claude weekly limit ran out, so switch models and keep going
No. The weekly and 5-hour limits are shared across all models, and switching does not restore them. Only when the message names the Opus limit or the Sonnet limit does /model, switched to the other family, keep you working.
- Keep the Codex banked reset for later; there is no hurry
Not quite. Banked resets expire, usually 30 days after they are granted; the expiry shown in your account is the one that counts, and an expired reset is not replaced. A reset is not consumed when there is nothing to refresh. So when one is close to expiring and both bars are well used, use it.
- Buying a Codex reset or using a banked one leaves the weekly reset time unchanged
No, it moves. After the reset, the next weekly reset is 7 days after the first message you send, and the old time no longer applies. A purchased reset applies as soon as you pay and cannot be saved for later. Claude is the opposite: the weekly reset time is fixed, and buying extra usage does not change it.
Tips
- Read the numbers first: /usage in Claude Code, /status in Codex. Compare the percentage used with the days left until the reset; the weekly limit calculator on this site tells you whether your current pace will last.
- Cut subagents first. It is the step with the fastest effect.
- Lower the reasoning effort one step, and use a smaller model for routine work.
- Codex: if you hold a banked reset, use it, then send a message right away. The next weekly reset is 7 days from that message.
- Claude: once you are on extra usage, do not leave a session idle. The cache lasts five minutes, and the first message after a break re-reads the whole conversation.
- Move large jobs that can wait to after the reset. The Codex and Claude reset pages on this site show the reset time and whether a reset has been announced.