Agent Upstart

Cool things people are building with agents — and practical help getting the most out of them.

Playbooks · Coding tasks

ChatGPT and Codex share one token budget

Written November 8, 2026

From December, Plus, Pro, and Business draw chat and Codex from one token pool. Instant chat stays outside it. What to measure before it lands.

What changes in December

OpenAI's Help Center page on ChatGPT and Codex usage limits, updated October 31, states it plainly: starting in December, the usage allowance on Plus, Pro, and Business plans will be shared across ChatGPT and Codex, as OpenAI brings the agentic capabilities of Work into the main ChatGPT experience. Chat in the evening and Codex in the afternoon will draw on the same pool. The change covers those three plans only. The page says nothing about Free, Enterprise, or Edu plans, so do not assume it applies to them.

Usage will be measured in tokens, the units of information a model processes and generates. How much of the allowance you use depends on the model you choose and the task you ask it to do: more advanced models and complex tasks generally use more. One exemption is named. Instant chat will not draw from your usage, so quick questions to the fast model stay outside the meter. Your combined remaining usage and reset times will appear in Settings, then Usage. If you reach the limit, you can wait for it to reset, upgrade your plan, or buy credits to continue using eligible features.

What OpenAI has not published

Three numbers that would let you plan are missing. OpenAI has published no exact December date, no per-plan token figure, and no credit prices. Without them, nobody can calculate whether a combined pool stretches further than the separate allowances it replaces, or whether a heavy Codex week will leave enough for chat. OpenAI says it expects most users will not need to wait for their usage to reset more often than they do now. That is an expectation, not a promise, and it ships with no figures behind it. The Decoder, reporting the change on November 1, called it the end of all-you-can-prompt pricing. That framing is the Decoder's view, not OpenAI's words, but the mechanics behind it are OpenAI's own: one meter, two products.

Measure one week before December

Since no published figure tells you where you stand, the only reliable input is your own burn, measured combined. Pick one representative week and record three numbers each day: roughly how much you used ChatGPT, roughly how much you used Codex, and which models did the work. The split matters less than the total, because December bills the total. A week with a long agent session and a week of short questions are different weeks; measure the busier kind if your work varies, and note which one you measured.

Then read the result against your plan, not against a guess. If your combined week already brushes the limits you hit today on either product alone, a heavy Codex day will now eat the chat budget too, and the fix is a plan decision, not a settings tweak. The Pro tier picker covered in GPT-6.1 Sol, Ultrafast, and the new Pro tiers still applies, but feed it combined demand: chat plus Codex, not Codex alone. If your week fits comfortably inside what you use today, the change is mostly bookkeeping, and the setting to watch is Settings, then Usage, once the combined meter appears.

What to do now

Limits

Related playbooks

Sources

Back to: Coding playbooks hub · Also: GPT-6.1 Sol, Ultrafast, and the new Pro tiers · Stay inside the Muse Code subscription allowance · Claude Code subagent updates now cost 99% less · Plan allowances: what each agent plan lets you do