ChatGPT will share Codex's limit in December: what one Codex job uses today
From December, ChatGPT conversations other than Instant chat will share an allowance with Codex on Plus, Pro and Business. Three of our Codex jobs of about 2 million tokens each moved the Plus five-hour meter by 30 to 46 points, so watch the usage page when the change lands and keep quick questions on Instant chat.
The 30-second version
From December, ChatGPT chats other than Instant will share Codex’s allowance on Plus, Pro and Business; on Plus, each of three long Codex jobs of ours moved the five-hour meter by 30 to 46 points.
- What changes. ChatGPT conversations, except Instant chat, will count against the same allowance as Codex.
- Our Plus sample. Three jobs of about 2 million tokens each moved the five-hour meter by 30 to 46 points and the weekly meter by 5 to 7.
- Pro today. There is no five-hour window; on 25 October our Pro account logged about 7 million tokens while its weekly meter rose from 0 to 6%.
- What to do. In December, check Settings → Usage, and keep quick questions on Instant chat, which won’t count.
What OpenAI announced
OpenAI’s help center says: “Starting in December, Plus, Pro, and Business plans’ usage allowance will be shared across ChatGPT and Codex.” Usage will be “measured in tokens”, and how much a request uses depends on the model and the task. “Instant chat will not draw from your usage.” Combined remaining usage and reset times will appear in Settings → Usage; at the limit, you can wait for the reset, upgrade or buy credits. OpenAI says it expects “most users won’t need to wait for their usage to reset more often than they do now.”
Codex’s pricing page already says “ChatGPT Work and Codex share usage” and “Pro plans currently have no five-hour limit”. Claude.ai and Claude Code also share usage limits.
What one Codex job uses today
These are four of our October Codex sessions. About nine in ten of their tokens were cached input.
| Plus job | Tokens | Five-hour window | Weekly limit |
|---|---|---|---|
| 17 Oct | 2.46 million | 11% → 57% | 24% → 31% |
| 25 Oct | 2.13 million | 6% → 43% | 1% → 7% |
| 25 Oct | 1.78 million | 48% → 78% | 7% → 12% |
| 31 Oct | 6.44 million | 0% → 86% | 9% → 22% |
Dividing each job’s tokens by its meter increase gives about 53,000 to 75,000 tokens per five-hour percentage point and 350,000 to 500,000 per weekly percentage point.
Our Pro account has only a weekly window. On 25 October it logged about 7 million tokens across nine sessions, and its weekly meter rose from 0 to 6%.
How we measured
- Data. Codex CLI session logs on one Mac. Each
token_countevent records the session’s token totals and the rate limits the server returned: plan, window length and percent used. - Accounts. One Plus account with five-hour and weekly windows; one Pro account logged as “prolite” with a weekly window; October–November 2026. The four Plus sessions used GPT-6 Astra at max effort.
- Points. The change in percent used between a session’s first and last event. The four Plus sessions neither overlapped other logged sessions nor crossed a reset.
What we didn’t test
- ChatGPT chats. They don’t count against Codex yet, and we can’t see their tokens.
- December’s weighting. Whether OpenAI counts cached, new and output tokens differently, or changes the allowances.
- Precision. The meter reports whole percentages, and use of the same accounts outside these logs would also move it.
Companion reading
- Codex vs Claude Code usage limits — the limit machinery on both sides, and why the wall arrives sooner on one.
- Claude Code usage limits — how the 5-hour and weekly limits work on Claude’s side.
- What agent traffic actually costs after Opus 5.5 — why cached tokens make long agent jobs cheaper than their size suggests.
Sources
- OpenAI Help Center — Understanding ChatGPT and Codex usage limits. Retrieved 4 November 2026.
- OpenAI — Codex pricing. Retrieved 4 November 2026.
- Hacker News — “ChatGPT and Codex usage limits merging”, 31 October 2026.
FAQ
When do ChatGPT and Codex start sharing usage limits? OpenAI’s help center says that starting in December 2026, the usage allowance on Plus, Pro and Business plans will be shared across ChatGPT and Codex. Usage will be measured in tokens, and Instant chat will not draw from it.
How much of a ChatGPT Plus Codex limit does one task use? In our October logs, three Codex jobs of about 2 million tokens each, on GPT-6 Astra at max effort, moved the Plus five-hour meter by 30 to 46 points and the weekly meter by 5 to 7. Most of those tokens were cached input.
Does ChatGPT Pro have a five-hour Codex limit? Not at the moment. OpenAI’s Codex pricing page says Pro plans currently have no five-hour limit, and our Pro account’s Codex logs show only a weekly window.
How can I make the shared allowance last? Use Instant chat for quick questions, since OpenAI says it won’t draw from your usage, and check Settings → Usage, where OpenAI says your combined remaining usage and reset times will appear.
Was this helpful?