If you pay for ChatGPT Plus, Pro, or Business and also use Codex, your limits are about to work differently. OpenAI says that starting in December, one token-based allowance will cover both ChatGPT and Codex. Heavy coding in Codex will leave less room for ChatGPT, and the reverse is also true. Only Instant chat stays outside the meter.

Coffee summary

  • FACT: From December, Plus, Pro, and Business plans will share one usage allowance across ChatGPT and Codex, according to an OpenAI Help Center article updated on October 31, 2026.
  • FACT: Usage will be measured in tokens, not messages. Instant chat will not draw from the allowance.
  • FACT: OpenAI has not published an exact start date or the size of any plan’s allowance.
  • CLAIM: OpenAI says it expects most users won’t need to wait for resets more often than they do now.
  • OPINION: The model picker becomes your main cost control.

What happened

OpenAI’s new help article, “Understanding ChatGPT and Codex usage limits,” says the change comes “as we bring the agentic capabilities of Work into the main ChatGPT experience.” Work is ChatGPT’s agent for longer tasks. Codex is OpenAI’s coding agent.

Part of this pool already exists. OpenAI’s Codex documentation says Codex, ChatGPT Work, ChatGPT for Excel, and Workspace Agents already share an allowance on plans that use shared allowances. The December change adds regular ChatGPT conversations to that pool, except Instant chat.

The news first surfaced on October 29, when ChatGPT’s web checkout began showing a line about a shared limit across ChatGPT and Codex starting in December, as reported by SaasBuzz. The help article put it in writing two days later.

Why it matters

Until now, many people treated chat and coding as separate budgets. After the change, a long afternoon of Codex work and an evening of research in ChatGPT will spend from the same balance.

Tokens are the small chunks of text a model reads and writes. Counting tokens means a long agent run costs more than a quick question, instead of both counting as “one message.” OpenAI says more advanced models and complex tasks “generally use more of your allowance.”

CLAIM: OpenAI’s own estimates show how much the model choice matters. On Plus, OpenAI lists roughly 5 to 45 local messages per five-hour window with GPT-6 Astra, its most capable model, versus 350 to 3,000 with the lighter GPT-6 Luna, as compiled by SaasBuzz from OpenAI’s model guide. OpenAI calls these ranges estimates, not guarantees.

What changes, at a glance

Today From December
Codex and Work Shared allowance Same shared allowance
Regular ChatGPT chat Separate model limits Draws from the shared allowance
Instant chat Separate Does not count
Unit Mixed, often messages Tokens
Where to check Varies by feature Settings → Usage, one combined view

Who should care

  • Developers on Plus or Business who use Codex for hours and ChatGPT for everything else.
  • Pro subscribers, who just received a bigger Work and Codex allowance on October 29. That larger pool will now also pay for chat.
  • Team leads budgeting Business seats, since how far a seat goes will depend on which models people pick.

Limitations and open questions

  • No date yet. OpenAI says “December,” nothing more specific.
  • No sizes. OpenAI has not said how many tokens any plan includes, or whether allowances will grow to absorb chat.
  • Windows unclear. It is not yet clear whether the current five-hour and weekly windows will stay the same.
  • “Instant” is undefined. OpenAI has not listed exactly which models count as Instant chat.

OPINION: This follows a pattern. Anthropic already shares plan limits across Claude and Claude Code, according to its help page as quoted by SaasBuzz. Shared pools are becoming the norm for AI subscriptions.

What to do next

  1. Check Settings → Usage now so you know your normal pattern before December.
  2. Use Instant chat for quick questions. It will not draw from the shared allowance.
  3. Match the model to the task. Save the heaviest models for hard problems and use lighter ones for routine work.
  4. Keep long Codex runs focused. Clear, scoped tasks use fewer tokens than open-ended ones.
  5. Know your fallback. At the limit, OpenAI says you can wait for a reset, upgrade, or buy credits for eligible features.

AIImpish Take

One shared allowance is simpler to understand but easier to drain. If you code in Codex every day, December is when model choice starts to matter for your chat time too. Watch OpenAI’s help article for the exact date and allowance sizes before you change plans.

Related on AIImpish

*Sources: OpenAI Help Center, Understanding ChatGPT and Codex usage limits; OpenAI Help Center, Using Codex with your ChatGPT plan; SaasBuzz, OpenAI puts ChatGPT and Codex on one token allowance in December.*