ChatGPT and Codex will share one usage allowance from December: what changes for developers

OpenAI has updated its help article on ChatGPT and Codex usage limits with a short notice: "Starting in December, Plus, Pro, and Business plans' usage allowance will be shared across ChatGPT and Codex as we bring the agentic capabilities of Work into the main ChatGPT experience." The New Stack reports the article was published quietly over the weekend, and Unite.AI covered the same notice on Monday.

For anyone who uses Codex as their main coding agent on a ChatGPT subscription, this is a change in how the budget works rather than a new feature. Coding sessions and chat conversations will draw from the same pool.

What OpenAI has confirmed

The notice on the OpenAI Help Center is brief. It says:

• Who: Plus, Pro and Business plans. Enterprise and Edu are not mentioned.
• When: "starting in December". No exact date is given.
• How it is counted: usage for ChatGPT and Codex will be measured in tokens. How much of the allowance a request uses depends on the model and the task, and "more advanced models and complex tasks generally use more of your allowance."
• What is exempt: "Instant chat will not draw from your usage."
• Where to check: combined remaining usage and reset times will be shown in Settings → Usage.
• At the limit: you can wait for the reset, upgrade your plan, or buy credits for eligible features.

OpenAI adds that it expects "most users won't need to wait for their usage to reset more often than they do now." It has not published the size of the combined allowance per plan.

What is already shared today

The change extends a pooling that already exists. According to OpenAI's Codex pricing page, "ChatGPT Work and Codex share usage", and usage limits are also shared with other agentic features once their pricing takes effect, which currently includes ChatGPT for Excel on Plus and Pro. A separate help article on managing usage with GPT-6 Astra describes the same shared Work and Codex allowance, with a five-hour window and a weekly window on some plans.

What changes in December is the boundary between that agentic pool and regular ChatGPT conversations. As The New Stack puts it, intensive Codex sessions will eat into the allowance available for ChatGPT conversations, and the other way around. The exception is Instant chat.

The current Codex documentation gives an idea of how uneven consumption already is. For Plus and Standard Business seats, OpenAI's pricing page estimates 5 to 45 local messages per five-hour window with GPT-6 Astra, 15 to 160 with GPT-6.1 Sol and 350 to 3,000 with GPT-6 Luna. Pro plans currently have no five-hour limit. Fast mode consumes included usage at 2.5 times the Standard rate. These are OpenAI's own estimates, not guaranteed counts, and they describe today's system, not the December one.

Why subscribers are pushing back

The notice follows a turbulent month for these limits. On September 29, OpenAI reopened Pro 200 subscriptions at $200 a month and said included usage allowances were changing. On October 29, the release notes announced that the Pro 200 reduction was cancelled, and that Pro 100 and Pro 500 plans now include twice their previous Work and Codex allowance.

According to The New Stack, a thread opened on OpenAI's developer forum on Saturday asks the company to keep the two allowances separate, or to guarantee that subscribers will not lose usable capacity. One user quoted by the publication wants a real-time dashboard showing consumption "by product, model, and task" before the change takes effect. Another points out that doubling an allowance while adding more things that consume it does not necessarily mean more usable capacity.

The New Stack says it asked OpenAI for the exact December date and whether the combined allowance would match the current separate ones. OpenAI declined to answer and pointed to its recent limit increases and existing documentation.

OpenAI is not alone in merging chat and agent work. The same article notes that Anthropic combined Claude Chat and Cowork into a single interface in September, although the report does not describe any change to Claude's usage limits.

What this means if you build with Codex

If you rely on a ChatGPT subscription for coding rather than an API key, a few practical points follow from the documentation:

1. Measure your usage before December. The Codex CLI shows remaining limits with /status, and the usage dashboard shows current limits and reset times, according to the pricing page. A week or two of data gives you a baseline to compare against once the pool is shared.
2. Use Instant for quick questions. Since Instant chat will not draw from the shared allowance, routine lookups and short questions can stay outside the budget your coding agent needs.
3. Pick the model per task. The spread between GPT-6 Luna and GPT-6 Astra in OpenAI's own estimates is large. Reserving the heaviest model for the hardest tasks will matter more when every conversation counts against the same pool.
4. Trim context. OpenAI's documentation recommends smaller prompts, fewer source files, a smaller AGENTS.md and fewer active MCP servers, since every MCP server adds context to each message.
5. Keep an API key path for automation. The pricing page says Codex can run with an API key in the CLI, SDK or IDE extension, billed at API rates and separate from subscription usage. For CI jobs or scheduled runs, that keeps your interactive allowance for interactive work.

What is still unknown

Three questions remain open until OpenAI publishes more detail: the exact start date in December, the size of the combined allowance on each plan, and whether the five-hour and weekly windows described in the current Codex documentation will still apply in the same way. OpenAI's statement that most users should not hit resets more often than today is a forecast, not a published quota.

We will update this article when OpenAI publishes the new limits.