OpenAI Will Merge ChatGPT's Chat and Work Tabs and Pool Usage With Codex

OpenAI is preparing to fold the Chat and Work tabs in ChatGPT into a single experience, and at the same time it is changing how paid plans meter usage. Starting in December, Plus, Pro, and Business subscribers will draw from one shared token allowance across ChatGPT and Codex, rather than separate buckets for each product.

The company told users by email that the combined experience will first go to a small group for feedback before a wider rollout. Codex, its software development agent, will stay a separate view inside the ChatGPT desktop app. But because the usage pool is shared, how much you code in Codex will now affect how much agentic work you can do in ChatGPT, and the other way around.

For developers who live in Codex and also lean on ChatGPT for research, documents, or quick prototypes, this is the most significant change to the subscription math since Work launched. Here is what is changing, what stays the same, and how to plan around it.

Key facts at a glance

• One interface: Chat and Work will merge into a single ChatGPT experience, tested first with a small group of users before a broad rollout.
• Codex stays separate: Codex remains its own view in the desktop app, with its own history.
• Shared allowance from December: Plus, Pro, and Business usage will be counted in tokens from a single pool covering both ChatGPT and Codex.
• Instant chat is exempt: conversations with Instant models do not draw from the shared allowance.
• Pro limits doubled: OpenAI says it doubled Work and Codex allowances on its Pro $100, $200, and $500 tiers, effective October 29.

Context: how ChatGPT got so complicated

Today, the ChatGPT desktop app has two top-level sections, ChatGPT and Codex. Inside ChatGPT there is a toggle between Chat, meant for fast conversational help, and Work, an agent built for longer, multi-step tasks and finished deliverables such as documents, spreadsheets, presentations, reports, and websites. Codex is dedicated to writing and debugging code, running tests and commands, reviewing changes, and working inside repositories.

That layout has drawn criticism for being hard to follow. By one count, users currently pick between two operating modes and 11 model settings. OpenAI's own help documentation lists separate models for Work and Codex, including GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Sol, and GPT-6 Luna, which are not available in regular chats, plus GPT-6 Astra for higher tiers.

Rival Anthropic took a similar step in September, merging Claude's chat and Cowork functions into one screen. OpenAI's move brings ChatGPT in line with that approach.

What the merged ChatGPT will do

In its email to users, OpenAI described the goal as letting people "move from exploring an idea to having ChatGPT carry it through, all in one place." In practice, that means Work's agentic tools appear inside the normal chat flow. You will be able to bring in files and plugins and ask ChatGPT to:

• Build a website or a simple app prototype.
• Create a presentation or other finished document.
• Do things on the web on your behalf, while you review and guide its work.

The company said: "Codex will remain a separate experience in the ChatGPT desktop app." So for code that lives in a repository, with a terminal, tests, and diffs, Codex remains the dedicated tool. The merged ChatGPT is aimed at everything around the code: research, specs, documentation, slides, and quick front-end mockups.

How the shared usage pool works

This is the part developers should pay closest attention to. According to OpenAI's support documentation, from December, Plus, Pro, and Business usage will be shared across ChatGPT and Codex and measured in tokens. Tokens measure the information a model reads and writes, so heavier models, larger context, and higher reasoning settings use more of your allowance per task.

A few details matter:

• Instant chats are free of the meter. Chats with Instant models do not draw from usage, so basic Q&A stays effectively flat-rate. OpenAI also says "Unlimited Instant" access is excluded from the shared pool.
• You can check your balance. Remaining allowance and reset times are shown under Settings > Usage.
• When you run out, you can wait for a reset, upgrade your plan, or buy additional credits.
• Model choice changes your burn rate. OpenAI notes that Astra may use your allowance faster than older Sol models, depending on task, input and output size, reasoning level, and speed.
• Long context costs nothing extra up front, but still counts. GPT-6.1 Sol has a default 500,000-token context window in Work and Codex, with no long-context surcharge inside the included allowance. Longer conversations still count toward limits.

OpenAI says it does not expect most users to hit resets more often. It also said that, depending on usage and model, some people may need to wait for resets more frequently.

The Ultrafast catch on the top tier

At launch, Ultrafast, a lower-latency mode for GPT-6 Astra and GPT-6.1 Sol, is limited to Pro $500 and eligible Enterprise and Edu workspaces. On Pro $500, Ultrafast consumes included usage at 8 times the Standard rate for the selected model, and after that allowance is used, it draws credits at 6 times the Standard credit rate. Ultrafast is not available on Plus, Pro $100, Pro $200, or Business, even with credits.

For a developer, that means one long Ultrafast session in Codex could now eat into the same pool you would use for a big Work research task later that day.

What it means for you

If you mainly use ChatGPT for quick questions, very little changes. Instant chats stay outside the meter, and the merged interface should be easier to navigate.

If you are a heavy Codex user on Plus, plan for the shared pool. A day of refactoring with a top model could leave less room for agentic ChatGPT tasks such as building a slide deck or running a web task. Practical steps:

1. Check Settings > Usage for a week before December to see how much Codex alone uses.
2. Match the model to the job. Use lighter models or lower reasoning for routine edits, and save Astra for hard problems.
3. Keep simple questions in Instant chat, which does not touch the allowance.
4. Watch context size. Long threads with large files count toward limits even without a surcharge.
5. Decide whether credits or an upgrade is cheaper for your pattern of use.

For Business admins, Work Cloud, Work Local, and Codex Local are controlled separately under workspace permissions, and starting defaults for Work and Codex models are set apart from Chat. Review those settings before the pooled meter takes effect so that teams are not surprised by resets.

The counterpoint

Some observers read the merger as the end of "all-you-can-prompt" pricing. On paper, ChatGPT remains flat-rate because Instant chats are exempt. In reality, the most capable features, the ones that justify paying for Plus or Pro, now sit behind a token budget, and heavy users on cheaper plans may find themselves pushed toward lighter models and lower-quality results. The timing of the Pro limit increase, just before the shared pool arrives, suggests OpenAI is trying to soften that shift for its highest-paying customers.

There is also a usability tradeoff. Folding Work into Chat could make it less obvious when a request is about to trigger a long, allowance-heavy agent run rather than a quick reply.

What happens next

The merged Chat and Work experience is in a limited test now, with a broader rollout to follow once OpenAI collects feedback. The shared ChatGPT and Codex allowance starts in December for Plus, Pro, and Business. Make sure the desktop app is up to date, since some models, including Astra, require the latest app and, for the command-line tool, Codex CLI version 0.153.0 or newer.

Conclusion

OpenAI is simplifying ChatGPT's interface while making its usage model more like a metered cloud service.

• Chat and Work become one, with files, plugins, web tasks, and deliverables in a single flow.
• Codex stays separate, but shares a token allowance with ChatGPT from December.
• Instant chats remain unmetered, so casual use is largely unaffected.
• Heavy coders should audit usage now, choose models carefully, and budget for credits or upgrades.

FAQ

Will Codex disappear into ChatGPT?

No. OpenAI says Codex remains a separate experience in the desktop app with its own history.

Do regular chats count toward the new limit?

Chats with Instant models do not. Agentic Work tasks and Codex sessions do.

Where can I see how much I have left?

Under Settings > Usage, which shows your remaining allowance and reset times.