What Are the Codex Usage Limits on Plus and Pro Now?
ChatGPT Plus ($20/mo) limits Codex with a five-hour window and a weekly allowance, worth roughly 5–45 GPT-6 Astra messages per five hours by OpenAI's estimate. The three Pro plans ($100, $200 and $500/mo) drop the five-hour window and keep a larger allowance. Type /status in Codex CLI to see what you have left.
Codex CLI also runs on a fixed monthly API with no five-hour or weekly window. Plans start at $19 a month with $20 of compute, and your ChatGPT allowance is still there when it resets. Set up Codex CLI →
Codex limits by plan (October 2026)
| Plan | Codex limits |
|---|---|
| Free / Go ($0 / $8) | GPT-6 Luna, desktop app |
| Plus ($20) | Five-hour + weekly |
| Pro 100 ($100) | No five-hour window |
| Pro 200 ($200) | Same, larger allowance |
| Pro 500 ($500) | Same, largest + Ultrafast |
| Business ($20/seat) | Standard: as Plus. Premium: 5x, no five-hour |
| Enterprise (custom) | Credits, or as Plus |
Monthly prices. Business is $20 per user billed annually, $25 monthly, from 2 users; Premium seats get 5x the Standard allowance. Source: OpenAI Codex pricing docs and Help Center, checked 2026-10-13.
How many Codex messages Plus gets per five hours
| Model | Per 5 hours |
|---|---|
| GPT-6 Astra | 5–45 |
| GPT-6.1 Sol | 15–160 |
| GPT-6 Sol | 15–150 |
| GPT-6 Luna | 350–3,000 |
Local messages. Standard Business seats get the same ranges; cloud tasks use more.
What changed on September 29, 2026
- OpenAI launched Pro 500 at $500/mo, the only Pro plan with Ultrafast mode for GPT-6 Astra and GPT-6.1 Sol.
- Pro 200 reopened to new subscribers at the same price with a smaller included allowance.
- Anyone with an active Pro 200 subscription between September 22 and 10 a.m. PT on September 29 keeps the old allowance through October 29, 2026.
- OpenAI has not published the new Pro 200 size. According to emails subscribers have posted, it drops from 20x to 10x the Plus allowance on October 30 and comes with a one-time grant of 62,500 credits that expires December 31.
What changed October 6 to 14, 2026
- October 6: OpenAI cut its API usage tiers from five to three (changelog). If you run Codex on an API key, Build starts at $5 of credit purchases with a $500 monthly usage limit, Launch at $100 with $5,000 and Grow at $500 with $200,000. GPT-6.1 Sol gets 1M, 4M and 40M tokens per minute on those tiers (rate limits).
- October 7: Codex CLI 0.161.0 made GPT-6.1 Sol the default model (release). On Plus that is 15–160 local messages per five hours, against 5–45 on GPT-6 Astra. On an API key it lists at $2 input, $0.10 cached and $10 output per million tokens.
- October 8: Codex CLI 0.162.0 lets custom Responses-compatible providers turn on web access and remote compaction (release), and the API's Ultrafast tier now covers GPT-6.1 Sol. In Codex, Ultrafast uses your allowance 8x faster.
- October 14: GPT-5.5 leaves ChatGPT and Codex (OpenAI). A config.toml that still names it needs another model.
- If you had no model set and now run out sooner, set model = "gpt-6-luna" in ~/.codex/config.toml for routine work. If the $500 Build limit stops an API-key setup mid-month, buying credits moves you up a tier.
How the five-hour and weekly limits work
- On Plus, you need room in both limits to keep working, so you can run out of the five-hour window with weekly allowance left.
- A five-hour window opens with your first message after the previous one closed. Heavy sessions can use it up in less than five hours.
- The weekly allowance runs on a seven-day cycle. Buying an instant reset starts a new week with your next request.
- Codex CLI, the IDE extension, the desktop app, web, cloud tasks and ChatGPT Work all draw from one allowance.
- Fast mode uses it 2.5x faster, Ultrafast 8x, and image generation roughly 3–5x.
- If you run out mid-task, Codex finishes the turn already running.
How to check your remaining Codex usage
- In Codex CLI, type /status during a session.
- On the web, open chatgpt.com/codex/settings/usage.
- In the ChatGPT desktop app, open Settings → Usage to see reset times, banked resets and your credit balance.
What to do when you hit the limit, cheapest first
- Wait for the reset time, or apply a banked reset if you have one.
- Stretch the next window: GPT-6 Luna or GPT-6.1 Sol for routine work, lower reasoning effort, Fast mode off, fewer MCP servers. Switching models after you run out does nothing.
- Buy credits. Plus and Pro accounts can keep working without changing plans, and some can buy an instant reset.
- Upgrade to Pro only if you run out most days.
- Move volume work to an API key or another provider and keep your plan for the work that needs it.
The catch
These are estimates OpenAI can change at any time: in September 2026 alone it launched a $500 tier and cut the Pro 200 allowance for new subscribers. Your own Settings → Usage page is the only number that counts for your account. If a workflow depends on a guaranteed volume, a windowed subscription can't promise it; a metered API or a budgeted flat plan can.
Where Standard Compute fits
Codex CLI accepts any OpenAI-compatible provider through [model_providers] in ~/.codex/config.toml, and it runs against Standard Compute's API at https://api.stdcmpt.com/v1. One flat plan from $19/mo gives you a monthly compute budget with smart routing and no five-hour or weekly windows. Run standardcompute launch codex to connect it for a session; those runs use your Standard Compute budget and leave your ChatGPT allowance alone.
FAQ
What does ChatGPT Pro get you in Codex now?
ChatGPT Pro gets you Codex with no five-hour window and a larger included allowance, and the $500 plan adds Ultrafast mode. OpenAI ranks the allowances Pro 100, then Pro 200, then Pro 500, but doesn't publish the exact sizes. New Pro 200 subscribers since September 29, 2026 get less than the old plan offered. When the allowance runs out, you can buy credits instead of upgrading.
How long until Codex limits reset?
/status in Codex CLI or Settings → Usage shows the exact time. On Plus, the five-hour window reopens with your first message after it closes, so there is no fixed clock time, and the weekly allowance resets every seven days. OpenAI Support can't reset limits for you.
Why did I hit the limit after a few messages?
Usually the model and settings. GPT-6 Astra can mean as few as 5 local messages per window on Plus, Fast mode spends the allowance 2.5x faster, and cloud tasks cost more than local ones. One visible agent message can also carry a lot of context, tool calls and MCP output.
Is Codex Pro worth $200?
Only if you run out on Plus most days. Pro 100 already removes the five-hour window, and new Pro 200 subscriptions now include less than the old 20x deal. If you run out only now and then, credits cost less than a plan jump.
What can I set up over lunch when Codex's weekly limit runs out mid-sprint?
Point Codex CLI at a flat monthly API: run standardcompute launch codex and Codex uses Standard Compute's Responses API, with GPT-5.6 Sol, Terra and Luna, GPT-6 Astra and GPT-6 Luna on one key and no 5-hour or weekly window. Plans start at $19 a month with a $20 compute budget; a sprint of heavy Codex work wants Standard ($89, $95 budget) or Pro ($249, $269 budget). At the budget, new requests get HTTP 402 instead of a bill, and your ChatGPT plan's Codex allowance is still there when it resets.