ChatGPT + Codex Will Share One Usage Allowance in December: Keep Your Coding Time From Eating Your Chat Time

In one sentence: starting in December, if you pay for ChatGPT Plus, Pro, or Business, your everyday chats and your Codex coding work will come out of one shared pot of usage — instead of two separate pots — and the only thing OpenAI says stays outside that pot is quick Instant chat.

Think of it like a household data plan. Today, your phone calls and your home internet each have their own limit. From December, OpenAI is putting ChatGPT chat and Codex on one family plan: a long afternoon of coding can leave less room for chatting, and a long afternoon of chatting can leave less room for coding. The price of the plan itself is not changing in this announcement — Plus is still $20/month and Pro is still $100/$200/$500 as currently listed — but how fast you burn through your included usage now depends on both activities together.

Two terms, plainly. A token is the unit OpenAI counts: roughly, the small chunks of words a model reads and writes. You never buy tokens one by one on a subscription — you just need to know that longer, harder tasks burn more of them. Codex is OpenAI’s coding assistant: it writes, edits, and runs code for you, inside ChatGPT Work, its own app, your code editor, or the web. Instant chat is ChatGPT’s fast, everyday chat mode for quick questions — and it is the one thing OpenAI explicitly says will not draw from the shared allowance.

What OpenAI actually announced

This is not a blog-post launch. It is a Help Center article, “Understanding ChatGPT and Codex usage limits,” updated October 31, 2026, and confirmed by multiple independent reads of that page on November 1–2. That sourcing matters: OpenAI has published no blog post about it, no exact December date, and no per-plan token figures. Here is what the article does say, quoted closely:

• Who: Plus, Pro, and Business plans.
• When: “Starting in December” — no day given.
• What: “Your plan’s usage allowance will be shared across ChatGPT and Codex” as OpenAI brings Work’s agentic capabilities into the main ChatGPT experience.
• How it is counted: in tokens. “More advanced models and complex tasks generally use more of your allowance.”
• The exception: “Instant chat will not draw from your usage.”
• Where to check: Settings → Usage will show your combined remaining usage and reset times.
• When you hit the limit: wait for the reset, upgrade your plan, or buy credits to keep going on eligible features.
• OpenAI’s expectation: “We expect most users won’t need to wait for their usage to reset more often than they do now.” That is the vendor’s forecast, not a measured result — treat it that way.

What is still unpublished (do not guess these)

• The size of the shared allowance for any plan — so nobody can honestly tell you yet whether the shared pool is bigger or smaller than the two separate limits it replaces.
• The exact December start date.
• Whether Free, Go, Enterprise, or Edu plans are affected at all — they are not named.
• Whether per-model message caps in chat stay alongside the new token pool, or go away.
• How image generation, file uploads, and regular voice chat fit into the new allowance — the article does not say.

If a guide or a salesperson gives you a token number for your plan before December, that number did not come from this announcement.

Why this matters even though prices are not changing

This change sits on top of an already-moving target. On October 29, OpenAI doubled the Work and Codex allowance for Pro 100 and Pro 500, and canceled a planned cut to Pro 200’s allowance — prices unchanged. December’s change is different: it merges the meters. For most people who only chat, or only code a little, nothing visible changes. The people who feel it are the ones who do both heavily: freelancers who draft proposals in chat all morning and run Codex all afternoon, small dev shops, and anyone selling Codex-based services to clients. For them, one long Codex run can now eat the chat allowance they were counting on for client communication later that day.

The 5-step prep checklist (free, 10 minutes, do it before December)

1. Open Settings → Usage in ChatGPT today. Write down two numbers: your current usage and your next reset time. This is your baseline — without it, December’s combined number means nothing.
2. Repeat once a week for the next two weeks. Note which weeks feel tight and what you were doing: heavy chat, heavy Codex, or both. The pattern, not any single day, tells you where the risk is.
3. Move your quick questions to Instant chat. Short, everyday questions — the kind you would ask in one sentence — belong on the fast mode that stays off the meter. Save the advanced models for work that actually needs them.
4. Batch your Codex runs. Instead of five scattered coding sessions a day, group related tasks into one or two focused sessions. Fewer restarts means less repeated context being re-read and re-billed against your allowance.
5. Re-check Settings → Usage in the first week of December. Compare the new combined remaining usage against your November baseline. If you are hitting the cap earlier, that is the week to decide — upgrade, buy credits, or split the work — not the week a client deadline lands.

The money play: sell a $149 Codex usage guardrail

Who buys it: freelancers, solo developers, and small agencies on Plus or Pro who use both ChatGPT and Codex and cannot afford to stall mid-project.

What you sell: a one-time setup, $149, plus an optional $49/month check-in. You do the 5-step checklist for them: baseline their Settings → Usage, label which of their recurring tasks belong on Instant chat vs. advanced models vs. Codex, write a one-page “when to use what” sheet for their team, and set a weekly usage-review reminder. The monthly check-in is you reading their usage trend and telling them, before they hit the cap, whether to upgrade, buy credits, or rebalance.

Why it is sellable now: the change is announced but the numbers are unpublished — that uncertainty is exactly what a busy freelancer will pay to have someone else watch. You are not selling access to a secret; you are selling the weekly check they will not do themselves.

How to validate in 7 days: offer the $149 setup to three ChatGPT-plus-Codex users you already know (a local dev, a freelancer group, a Discord or Slack community). If none of the three cares, the play is dead for your audience — the change only bites heavy dual users, and that filter is the point. If one buys, deliver the checklist, screenshot their baseline (with permission), and ask for a one-line testimonial about what the baseline revealed. Price note: the $149/$49 figures are this guide’s suggested starting prices, not OpenAI’s — OpenAI’s prices are the subscription prices above.

Pitfalls and limits

• Do not promise a client unlimited Codex runs on a Plus plan. Plus has a five-hour window and a weekly cap on Codex today, and December merges the meters further. Quote jobs with a usage buffer, not a “run it as much as you like” promise.
• Do not quote a token allowance for any plan yet. OpenAI has not published one. Anyone who does is guessing.
• Do not assume Free or Go users are affected. They are not named in the article. If you advise a Free or Go user, say so explicitly: “not covered by this announcement as of November 2.”
• Credits are not a magic refill for everything. The Help Center framing is that credits let you continue on eligible features — not that every feature accepts credits. Check the credits page for your plan before promising a client a workaround.
• This is a usage-accounting change, not a new model. If someone tries to sell you a December “upgrade” to get the shared allowance, that is not a thing — the allowance change applies to the plans named above automatically.

The bottom line

December does not raise your ChatGPT bill. It merges two meters into one, counts both in tokens, and leaves Instant chat outside the meter. The smart move costs nothing: baseline your usage now, route quick questions to Instant, batch Codex work, and re-check in December’s first week. If you sell coding help, that same checklist — packaged as a $149 setup — is a service clients will understand the moment their first shared-allowance warning appears.