OpenAI merges ChatGPT and Codex into one token pool

If you spend any serious time in ChatGPT's desktop app, you already know the UI has gotten messy. You have separate sections for ChatGPT and Codex, and inside the chat side, there are separate tabs for Chat and Work. Moving between a quick thought and an actual multi-step project usually meant jumping across screens and losing track of your files.

OpenAI just revealed that starting in December, Plus, Pro, and Business plans will see their ChatGPT and Codex usage limits merged into a single shared token allowance pool. Alongside that change, according to an email obtained by TechCrunch, OpenAI is rolling out an update to select testers that folds the Work tab directly into the main Chat window.

One window to brainstorm and build

OpenAI's goal here is to bake agentic tools right into the default chat surface instead of treating real work like a separate destination.

In the new setup, you start exploring an idea in chat, pull in files and plugins, and tell ChatGPT to run with it without hopping tabs. That means having it build a website, assemble a slide deck, or run tasks on the web while you review its output in real time. Codex will still stick around as a distinct experience inside the desktop app, but the actual computational limits running both tools are now tied together.

Shared tokens and a single usage meter

Under the hood, OpenAI is moving away from split quotas and switching entirely to a shared pool measured in tokens. What you burn will depend strictly on the model you call up and how heavy the prompt chain gets. Complex workflows and frontier models will chew through your allowance faster, while lightweight instant chat will not draw from the shared bucket at all.

OpenAI says most regular users will not see their daily routine interrupted or face more reset lockouts than before. Still, the company acknowledged that people leaning heavily into complex tasks might hit ceilings more often.

When the update lands in December, you will be able to track your shared allowance and countdown timers under Settings → Usage. If you run out of juice mid-project, you have three familiar choices: wait for the clock to reset, bump up your subscription tier, or pay out of pocket for add-on credits.

Why it matters

We think this is OpenAI finally cleaning up its fragmented product design to prepare for an agent-first world.

For power users, having agentic actions right where you already type is a major quality-of-life win. You no longer need to decide in advance whether a prompt is a simple question or a full project. The flip side is economic: pooling tokens across Codex and agentic chat means your developer workload and your deep research tasks now compete for the exact same bucket. Heavy users will feel the squeeze faster, which conveniently nudges them toward OpenAI's credit store.

It is also a direct reaction to the competition. Anthropic unified Claude's chat and Cowork functions into a single screen back in September. The battle among frontier labs is moving away from standalone chatbots and toward integrated workspaces. Unifying the limits is smart operational housekeeping for OpenAI, but users will have to watch their token consumption much more carefully once the switch flips.