“Unlimited” is rare in AI coding tools. Every one of them meters you somehow, and the method matters as much as the price: a plan that resets every five hours feels very different from one that hands you a monthly credit balance. There are four models in use today.
1. Rolling windows: Codex and Claude Code
Codex and Claude Code count usage in a rolling five-hour window plus a weekly cap, and the amount you get depends on the model you choose.
OpenAI publishes estimates rather than fixed numbers. On Plus and standard Business seats, it estimates 15–160 local messages per five hours with GPT-6.1 Sol, 35–285 with GPT-6.1 Terra and 350–3,000 with GPT-6 Luna. Cloud tasks use more than local ones. Codex Pro plans currently have no five-hour limit, though weekly limits can still apply.
Claude Code draws from the same pool as Claude chat on web, desktop and mobile. Max gives five times Pro’s allowance in both windows. Ultra gives ten times Pro’s weekly allowance and removes the five-hour window for verified accounts. Fable, Anthropic’s top model, has its own weekly limit on Max and Ultra.
Coming in December: OpenAI says ChatGPT and Codex will share a single token-based allowance on Plus, Pro and Business.
2. Monthly credits: Copilot and Kiro
Copilot now bills in GitHub AI Credits, worth one cent each. Pro includes 1,500 a month, Pro+ 7,000 and Max 20,000. Business and Enterprise seats add 1,900 and 3,900 credits per user to a shared pool. Inline completions and next edit suggestions don’t use credits on paid plans, and allowances reset on the 1st of each month.
Kiro works the same way with its own credits: 50 free, 1,000 on Pro, up to 10,000 on Power. Extra credits cost $0.04 each.
3. Weekly pool plus API pricing: Cursor
Cursor gives each plan a weekly pool for its own models (Grok and Cursor’s models), sized as a multiple of Plus: 4× on Premium, 15× on Super, 50× on Ultra. Third-party models such as Claude, GPT and Gemini sit outside the pool and are billed from the first request at API list price, from prepaid credit. Teams and Enterprise pay an extra $0.25 per million tokens on eligible third-party models.
4. Pay the provider directly: Cline, Zed, Amp
Tools such as Cline, Zed (Personal plan) and Amp let you use your own API keys or existing subscriptions, so your limit is whatever your model provider allows. See AI coding tools that work with your own API key.
What happens at the limit
| Tool | When you run out |
|---|---|
| Codex | Plus and Pro can buy credits; flexible-pricing workspaces buy workspace credits; anyone can continue locally with an API key at API rates |
| Claude Code | Wait for the reset, move up a plan, or turn on usage credits at standard API rates |
| Copilot | Upgrade, set a budget at $0.01 per credit, switch to a cheaper model or wait for the monthly reset (Free plans can’t buy extra) |
| Cursor | Prepaid credit top-ups, optional auto top-up and a monthly cap on extra usage |
| Kiro | Add-on credits at $0.04 each |
How to make an allowance last
- Pick the lighter model for routine work. Every rolling-window tool burns heavier models faster. In Codex, GPT-6.1 Terra and Luna stretch much further than Sol.
- Watch speed modes. Codex’s Fast mode uses included usage at 2.5× and Ultrafast at 8×.
- Prefer local over cloud tasks in Codex, since cloud tasks draw more.
- Check the usage page before big jobs. Both ChatGPT and Claude show remaining usage and reset times in Settings → Usage.
Frequently asked questions
What is the five-hour limit in Codex and Claude Code?
Both meter usage in a rolling five-hour session window, with a separate weekly limit on top. How much you get depends on the plan and the model: heavier models use the allowance faster. Codex Pro plans currently have no five-hour limit, and Claude's Ultra plan removes it for verified accounts.
How do GitHub Copilot AI credits work?
Copilot measures usage in GitHub AI Credits, where one credit is one US cent. Pro includes 1,500 credits a month, Pro+ 7,000 and Max 20,000. Inline completions and next edit suggestions don't use credits on paid plans. Allowances reset at 00:00 UTC on the 1st of each month.
What happens when I hit my limit?
It depends on the tool: Codex lets Plus and Pro users buy credits or continue with an API key; Claude lets you wait, upgrade or turn on usage credits at API rates; Copilot lets you set a budget at $0.01 per credit; Cursor sells prepaid credit top-ups.
Do ChatGPT and Codex share a usage limit?
OpenAI says that starting in December 2026, Plus, Pro and Business plans will share one token-based allowance across ChatGPT and Codex. Instant chat won't draw from it.