Connect with us

NEWS

OpenAI Folds ChatGPT Work Into Chat and One Token Pool

OpenAI is folding ChatGPT Work into ordinary chat for testers, then putting Plus, Pro and Business on one token pool with Codex in December.

Published

on

OpenAI is folding ChatGPT Work into ordinary chat for a test group, and from December Plus, Pro and Business will share one token pool with Codex.

A November 2 email to subscribers said a conversation will run from an idea to a finished job in one place, with files, plugins, websites, decks and web tasks in the same view. Codex stays a separate section in the ChatGPT desktop app.

The Work Tab Is Coming Into Chat

The desktop app still splits ChatGPT from Codex, then splits ChatGPT again into Chat and Work. That second split is the one OpenAI is taking down. The combined view goes first to select users for feedback. OpenAI has not named a date for everyone else.

This will enable you to move from exploring an idea to having ChatGPT carry it through, all in one place. You can bring in your files and plugins and ask ChatGPT to build a website, create a presentation, or do things for you on the web, while reviewing and guiding its work. Codex will remain a separate experience in the ChatGPT desktop app.

OpenAI, email to ChatGPT users, November 2, 2026

Work was never a second chatbot. It was the agent lane: longer jobs, connected apps, local files on desktop, and a built-in browser. Pulling that lane into Chat means a thread that started as a question can keep going until there is a site or a slide deck, without a tab change in the middle.

WHAT THE COMBINED VIEW CAN DO

  • Files and plugins: Bring documents and connected apps into the same conversation that started as a chat.
  • Websites and decks: Ask ChatGPT to build a site or a presentation without opening Work first.
  • Actions on the web: Hand it a job in the browser, then review and steer while it runs.
  • Codex aside: The coding agent keeps its own section in the desktop app, even after Chat and Work share a screen.

A combined Chat and Work view has already appeared on some accounts, which matches a limited test rather than a global flip. Users outside that group still see two tabs.

One Shared Pool in December

The interface change is the visible half. The meter is the other half. OpenAI’s help page says that starting in December, Plus, Pro and Business usage will be shared across ChatGPT and Codex as it brings the agent capabilities of Work into the main ChatGPT experience.

Usage will be counted in tokens, the units a model reads and writes. The model you pick and the size of the job decide how fast the bar moves. Advanced models and complex tasks use more. Instant chat will not draw from that allowance.

Work and Codex already share one plan allowance. A long research job in Work leaves less room for a coding session, and the reverse is also true. December pulls the rest of ChatGPT, Instant aside, into that same bar. OpenAI says it is preparing the change to limit disruption, and that it expects most people will not wait on resets more often than they do now.

Remaining usage and reset times will sit under Settings, then Usage. If the bar hits zero, the options are to wait, upgrade the plan, or buy credits for eligible features. Switching models does not restore a shared pool.

Anthropic Already Merged Cowork Into Claude

Anthropic moved first. On September 16 it wrote that Cowork and chat are merging into one Claude, so a quick question and a report due at noon live in the same conversation, even after the laptop closes.

The company had built Cowork as a separate place for bigger work, and Design for visual work. People used both, then got stuck choosing a tab, and work started in one place did not carry into the other. Anthropic stopped asking them to choose. Claude Docs and Claude Slides shipped the same day, and Claude Design began working inside ordinary conversations. Pro, Max and Ultra moved first on web, desktop and mobile. Team and Free were slated to follow. Enterprise admins were promised at least 30 days of notice. On October 23, Anthropic said the unified experience was generally available for Claude Enterprise.

OpenAI’s November note is the same product bet with a billing twist Anthropic already lives with: one front end for talk and agent work, and a coding agent that still has its own home. Claude Code stayed outside that merge. Codex is OpenAI’s version of that leftover room.

What Instant Chat Still Leaves Unlimited

The December pool is not a hard cap on every keystroke. Instant chat stays outside it. That is the lane OpenAI is leaving unmetered against the shared allowance, so a short back-and-forth does not have to compete with a Codex run.

Everything else in ChatGPT that burns tokens will. A long reasoning pass, a file-heavy job, a web task that used to live in Work, and a Codex session will draw from the same plan bucket on Plus, Pro and Business. The practical split is no longer Chat versus Work. It is Instant versus everything that actually spends.

That is why the tab merge and the pool landed in the same note. Once Work’s agent jobs sit in the main thread, OpenAI is no longer pretending those jobs are a side product with a side quota. They are ChatGPT, and they share a tank with the coding agent.

Plus Still Hits a Five-Hour Cap

The people who will feel the shared bar first are Plus and Standard Business users who already live inside a five-hour window. OpenAI does not publish a single message count. It publishes ranges, and it calls them estimates, not guarantees. Cloud tasks can cost more than local messages, and a weekly limit may sit on top of the five-hour one. You need allowance left in both to keep going.

On Plus and Standard Business, OpenAI estimates 5 to 45 local messages in five hours on GPT-6 Astra, 15 to 160 on GPT-6.1 Sol, 15 to 150 on GPT-6 Sol, and 350 to 3,000 on GPT-6 Luna.

FIVE-HOUR ESTIMATES ON PLUS

Model Plus and Standard Business, local messages
GPT-6 Astra 5 to 45
GPT-6.1 Sol 15 to 160
GPT-6 Sol 15 to 150
GPT-6 Luna 350 to 3,000

Pro 100, Pro 200 and Pro 500 currently have no five-hour usage limit in Work and Codex. They still have an included allowance, visible in the same Usage screen. On November 8, Tibo, who works on Codex and ChatGPT at OpenAI, wrote that all Pro plans skip the five-hour cap and that Pro 100 is 10X Plus, Pro 200 is 20X, and Pro 500 is 50X with Ultrafast.

Those multiples treat Plus as the unit the rest of the ladder is sold against. Plus did not get a matching lift on October 29, when Tibo said OpenAI was doubling Work and Codex limits on Pro 100, Pro 200, Pro 500 and Business Premium after dropping a planned cut to Pro 200. A heavy Codex day on Plus already eats the same window Work uses. From December, non-Instant chat joins that fight.

ChatGPT Work Arrived on July 9

The tab OpenAI is now folding up is four months old. On July 9, OpenAI launched ChatGPT Work as an agent that can take action across apps and files, stay with a project for hours if needed, and turn a goal into finished sheets, slides, docs and web apps. It shipped on GPT-5.6. Plugins reached Slack, Microsoft Teams, Google Drive, SharePoint, email, calendars and CRMs. Desktop added local files, a built-in browser and Computer Use, which can click, type and move files in the background.

OpenAI said more than 5 million people use Codex every week, and that more than 1 million of them now use it for work outside software. Work was the product shaped for that second group. Nearly 100% of teams inside OpenAI, including finance and sales, were already on ChatGPT Work and Codex, the company said. Web and mobile started with Pro, Enterprise and Edu, then Plus and Business. The desktop app put Chat, Work and Codex on every plan, including Free, for Mac and Windows. The old Codex app became that desktop app. Developers could still default to the Codex view and the Codex icon.

FROM TWO TABS TO ONE POOL

  1. July 9, 2026: ChatGPT Work launches as a separate agent tab beside Chat, with Codex still its own desktop mode.
  2. July 17, 2026: Tibo writes that the new desktop app did not land quite right, then adds Chat and Work switching and synced history across web, mobile and desktop.
  3. September 16, 2026: Anthropic merges Claude Cowork and chat into one Claude on Pro, Max and Ultra.
  4. October 29, 2026: OpenAI doubles Work and Codex limits on Pro 100, Pro 200, Pro 500 and Business Premium.
  5. November 2, 2026: An email says Chat and Work will combine for a test group, with Codex still separate on desktop.
  6. December 2026: Plus, Pro and Business start drawing ChatGPT and Codex from one token pool, Instant chat excluded.

The July 17 fix is the tell. OpenAI already knew the Chat and Work split confused people, and it tried to paper over it with a mode switch eight days after launch. The November email goes further: stop asking people to pick a tab at all.

Codex Stays Its Own Desktop Room

Codex is the piece OpenAI is not folding. The email is explicit. The coding agent remains a separate experience in the ChatGPT desktop app, with its own sidebar, diffs and cloud jobs. What it loses is a private quota. From December, a long Codex session and a long ChatGPT job are the same budget on Plus, Pro and Business.

That is a clean split on the screen and a messy one on the meter. Developers who live in Codex keep the layout they had. They do not keep a tank that chat cannot touch, unless they stay on Instant for everyday questions. API use is billed on its own, outside the subscription allowance, so a second path still exists for teams that want coding usage off the ChatGPT bar.

The test group will show whether one thread can go from a rough idea to a site or a deck without a second tab. The December pool will show whether that extra work still fits beside Codex in the same week.

Harry is the editor of SOMALI UPDATE, an independent title he owns and runs. Ten years in journalism, from reporter to editor, have settled into a set of verification habits he applies to every story. A quote is checked against the recording or transcript it came from. A statement attributed to an organisation is confirmed on that organisation's own channels before it is repeated. A figure is traced to the dataset or filing that first published it, and a photograph is checked for when and where it was actually taken. If any of those checks fails, the claim is left out or clearly marked as unconfirmed. Those habits cover the whole site, which reports news, business, technology, science and sports along with entertainment, lifestyle, travel, auto and gaming for readers around the world. Product claims in the technology, auto and gaming pages are tested in use where Harry can get his hands on the product. Corrections are published under a public policy and noted on the article. Readers who want to question a fact can write to support@somaliupdate.com.

Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Trending