Codex “You've hit your usage limit”: Fixes and Reset Times
You've hit your usage limit. Upgrade your plan or add credits to continue, or try again later.
It means your ChatGPT plan's Codex allowance is used up: wait for the reset, buy credits, or point Codex CLI at another provider to keep working now. On Plus that is the five-hour or weekly limit, on Pro the included allowance; type /status in Codex CLI or open Settings → Usage to see which one you hit and when it resets. If you would rather not upgrade to Pro, Codex CLI also runs on an API key: your own OpenAI key bills at per-token API rates, while a custom OpenAI-compatible provider such as Standard Compute (base URL https://api.stdcmpt.com/v1, from $19/month) gives Codex a fixed monthly bill with no five-hour window.
What causes it
- Plus and Standard Business seats have two limits, a five-hour window and a weekly allowance, and you need room in both. The window opens with your first message after the last one closed, so it can run out in less than five hours.
- Pro 100, Pro 200 and Pro 500 have no five-hour window but still cap total usage. New Pro 200 subscriptions since September 29, 2026 include less than before.
- Codex CLI, the IDE extension, the desktop app, web, cloud tasks and ChatGPT Work all draw from one allowance.
- Model and speed matter most: OpenAI estimates 5–45 local messages per window on GPT-6 Astra on Plus and 350–3,000 on GPT-6 Luna. Fast mode spends it 2.5x faster.
- Large repos, long sessions, many MCP servers and a long AGENTS.md add context to every message.
How to fix it
- Wait for the reset time shown by /status or at chatgpt.com/codex/settings/usage, or apply a banked reset if you have one.
- Buy credits. Plus and Pro accounts can keep working without changing plans.
- Stretch the next window: GPT-6 Luna or GPT-6.1 Sol for routine work, lower reasoning effort, Fast mode off. Switching models after you run out does nothing.
- Upgrade to any Pro plan ($100, $200 or $500/mo) to drop the five-hour window.
- Run Codex CLI on an API key, billed per token at API rates, or on another OpenAI-compatible provider.
Running an agent? Cloud tasks and parallel subagents multiply usage: three parallel tasks drain the shared allowance about three times as fast.
Compare APIs for Codex CLI
Best API for Codex CLI →
3 options compared, with prices and limits.
Keep working now: point Codex CLI at another provider
Codex CLI runs on any provider that serves the OpenAI Responses API: add a [model_providers] entry in ~/.codex/config.toml with its base URL and key (setup and trade-offs, providers compared on price). On Standard Compute, set base_url https://api.stdcmpt.com/v1 and model standardcompute, or run standardcompute launch codex for one session: each step goes to the cheapest model that can do it, on a flat monthly plan with no five-hour or weekly window, and your ChatGPT allowance stays untouched.
Plans from $19 a month with $20 of compute. Eligible new accounts can try $0.25 of compute first, no card. Trial terms
FAQ
Add credits or switch provider?
Add credits if you run out only now and then: Plus and Pro accounts keep working without changing plans, and credits cost less than a plan jump. Switch provider if you run out often or want a fixed bill: your own OpenAI API key bills per token, while a flat plan stops at its monthly budget instead of billing more.
Does the Codex limit reset on its own?
Yes. On Plus, the five-hour window reopens with your first message after it closes and the weekly allowance resets every seven days; /status shows the exact times. OpenAI Support can't reset limits for you.
Do Pro plans still have a five-hour Codex limit?
No. OpenAI's docs say Pro 100, Pro 200 and Pro 500 currently have no five-hour limit in Codex or ChatGPT Work, though each still has an included allowance.
Is API-key Codex unlimited?
It sits outside the plan allowance but bills per token at API rates, so heavy use gets expensive. A flat-rate OpenAI-compatible provider keeps the cost fixed within that plan's budget.