ChatGPT Work and Codex Now Share One Desktop App: What Actually Changed

TL;DR
OpenAI is consolidating its desktop apps, not merging ChatGPT and Codex into one indistinguishable product.
Official Sources#
| Source | Link |
|---|---|
| ChatGPT Work Announcement | openai.com/index/chatgpt-for-your-most-ambitious-work |
| GPT-5.6 Announcement | openai.com/index/gpt-5-6 |
| Codex GA Announcement | openai.com/index/codex-now-generally-available |
| ChatGPT Desktop Downloads | openai.com/chatgpt/download |
| GPT-5.6 API Pricing | platform.openai.com/docs/models/gpt-5-6 |
Last updated: July 15, 2026. Links verified against official OpenAI announcements.
OpenAI made a product consolidation today that is easy to describe badly. The Codex desktop app is becoming the new ChatGPT desktop app. That does not mean ChatGPT and Codex have merged into one product, that Codex is being shut down, or that every task should now use the same agent.
The useful way to read the announcement is simpler: one desktop shell now exposes three distinct modes - Chat, Work, and Codex. ChatGPT Work is for longer, cross-app knowledge work. Codex remains the coding agent for developers and technical professionals. GPT-5.6 is the new model family powering both surfaces at different capability tiers and effort settings.
That is a meaningful change for anyone who has been bouncing between a general assistant and a coding app. It is also a reason to get more explicit about task boundaries, permissions, and review.
What is actually merging#
OpenAI says that, starting July 9, the Codex app is merging with the new ChatGPT desktop app for Mac and Windows. Existing Codex app users update as usual and receive the new ChatGPT desktop app. The older ChatGPT desktop app is being renamed ChatGPT Classic.
The shared app is a distribution and workflow change. The Codex workspace is still present, and OpenAI explicitly says Codex remains its coding agent. In the updated desktop app, developers can set Codex as the default opening view and choose the Codex logo as the app icon. Desktop Codex projects also remain available from the ChatGPT mobile app.
So this is not a company merger, a model merger, or a retirement notice. It is one place to move from a question to an operational workflow or a repository.
| Mode | Best fit | What to expect |
|---|---|---|
| Chat | Questions, drafts, exploration, and quick analysis | A conversational assistant for interactive work |
| Work | Long-running work across connected apps, files, browser context, and artifacts | An agentic workflow you can inspect, steer, and approve |
| Codex | Repository-aware implementation, diffs, tests, and pull request review | A coding agent built around technical projects |
The distinction matters because a polished slide deck and a tested code change have different sources of truth. The shared desktop app removes app switching. It does not remove the need to choose the right control surface.
What ChatGPT Work adds#
ChatGPT Work is an agent inside ChatGPT for tasks that are larger than one prompt. OpenAI describes it as able to pull information from connected apps and workflows, create sheets, slides, documents, and web apps, and keep working on complex projects by breaking them into smaller steps.
Its core inputs are the places work already lives. The new plugins directory connects systems such as Slack, Microsoft Teams, Google Drive, SharePoint, email, calendars, CRMs, and project trackers. You can direct a prompt to a specific plugin with @, while ChatGPT can also suggest relevant connected tools.
On desktop, OpenAI is also adding a built-in browser, access to local files and apps, and Computer Use for actions across apps, tools, and the browser. Those capabilities should change how you frame a request: give the agent a concrete outcome, identify the approved sources, and state the human checkpoint. “Prepare a launch brief from these files and stop before sharing it” is a better operating instruction than “handle the launch.”
Scheduled Tasks can run once, on a schedule, when an event occurs, or while monitoring for changes. That makes Work suitable for recurring preparation and synthesis. It is not a reason to automate irreversible actions without review.
OpenAI is also introducing Sites in public beta, which can turn work into a shareable interactive site or web app. Treat it as an artifact workflow, not a substitute for production engineering.
Where Codex stays distinct#
Codex is still the developer mode. The general-availability announcement describes it as an agent that works in the editor, terminal, and cloud under a ChatGPT account, with SDK and Slack integration options for engineering teams. Its job is not merely to produce text about a change. It is to work through a technical task in the environment and return something reviewable.
Within the consolidated app, OpenAI highlights inline editing in diffs, pull request review in the side panel, faster computer use, and multiple repositories in one project. Codex still owns the implementation loop.
For engineers, the strongest workflow is usually a handoff between modes rather than a forced choice:
- Use Chat or Work to gather context, compare options, turn meetings or requirements into a concise technical brief, and identify dependencies.
- Move the scoped implementation into Codex, with repository instructions and a definition of done.
- Review the diff, run the relevant checks, and keep a human accountable for merge and release decisions.
Do not confuse this with asking Work to “build the app” and accepting the first artifact. Codex is the better mode when the task depends on a real codebase, local tooling, tests, or pull request review. Work is the better mode when the task depends on distributed business context and needs a document, plan, presentation, or connected-app workflow as its output.
GPT-5.6: the model layer under both products#
The app consolidation and the GPT-5.6 launch landed together, but they answer different questions. The desktop app tells you where to work. GPT-5.6 determines the capability and cost profile available in that work.
OpenAI is launching three generally available GPT-5.6 tiers:
| Tier | OpenAI's positioning | API price per 1M tokens |
|---|---|---|
| Sol | Flagship model | $5 input / $30 output |
| Terra | Balanced model for everyday work | $2.50 input / $15 output |
| Luna | Fastest and most cost-efficient model | $1 input / $6 output |
For ChatGPT Work and Codex, Free and Go users receive GPT-5.6 Terra. Plus, Pro, Business, and Enterprise users can choose Sol, Terra, or Luna and set an effort level. OpenAI says max is available to all users who can access GPT-5.6 in Work and Codex. ultra is available in Work for Pro and Enterprise users, and in Codex for Plus and higher plans.
The names are useful only if they guide a decision. Start with Terra for routine work that needs a capable default. Use Luna when speed and cost matter more than maximum reasoning. Escalate to Sol for difficult design, long-horizon reasoning, or demanding coding tasks. Then choose higher effort only after the task is properly scoped. More compute cannot rescue vague requirements or compensate for missing verification.
For API teams, GPT-5.6 adds Programmatic Tool Calling in the Responses API, letting the model write and run in-memory programs to coordinate tools and process intermediate results. The launch also introduces a multi-agent beta that can run concurrent subagents and synthesize their work in one request. These are API capabilities, not a promise that every desktop task is secretly running an arbitrary agent swarm.
A practical decision guide#
Choose ChatGPT Work when the work is spread across documents, browser research, connected apps, and recurring operational steps. Use it to assemble and refine a human-reviewable artifact. Grant only the app and file access the task actually needs, and keep approvals on for consequential actions.
Choose Codex when the output must be a reliable software change. Use it when you need repository context, a local or cloud development environment, diffs, tests, and pull request review. Keep instructions in the repository, define a narrow acceptance test, and inspect the result before merging.
Choose Chat when you need to think aloud, learn, write a first draft, or make a fast decision. It is often the right first stop, even if Work or Codex will take the next step.
The headline is not “one agent replaces every workflow.” It is that OpenAI now puts a general assistant, a cross-app work agent, and a coding agent under one desktop roof. That will make transitions faster. Good operators will still treat context, permissions, testing, and review as separate disciplines.
FAQ#
Is Codex being discontinued?#
No. OpenAI says Codex remains its coding agent for developers and technical professionals. The Codex app is becoming the new ChatGPT desktop app, where Codex is available alongside Chat and Work.
Is ChatGPT Work the same as Codex?#
No. ChatGPT Work is built for longer, cross-app workflows and shareable artifacts such as documents, slides, sheets, and sites. Codex is the coding mode for repository-aware implementation, diffs, testing, and pull request review.
Which GPT-5.6 model should developers use?#
Use Terra as a capable default, Luna when speed and cost are the priority, and Sol for demanding reasoning or coding work. OpenAI's available tiers and effort settings depend on the ChatGPT plan and product surface.
Does the new desktop app replace the existing ChatGPT desktop app?#
OpenAI says the existing ChatGPT desktop app will be renamed ChatGPT Classic. The updated app is available globally for Mac and Windows, with Chat, Work, and Codex on every plan, including Free.
Continue Reading#
- OpenAI Apps SDK: Building MCP UIs Inside ChatGPT
- ChatGPT Agent: OpenAI''s Operator Meets Deep Research
- ChatGPT Atlas: OpenAI''s Built-In Web Browser
- GPT-5.6 Sol Ultra Coming to Codex with Cooperative Subagents
- Codex Gets Computer Use in the EU - and a Clean Claude Code Import
- OpenAI Retunes GPT-5.6 Sol in ChatGPT and Makes Luna the Free Tier Default
Sources#
- ChatGPT is now a partner for your most ambitious work - OpenAI, July 9, 2026. Accessed July 9, 2026.
- GPT-5.6: Frontier intelligence that scales with your ambition - OpenAI, July 9, 2026. Accessed July 9, 2026.
- Codex is now generally available - OpenAI, October 6, 2025. Accessed July 9, 2026.
Get the next deep dive like this in your inbox
One email a week on OpenAI and the rest of the AI dev stack. Free.
Read next on AI coding tools
OpenAI Codex: Terminal and Cloud AI Coding Agent
Codex works from the terminal, cloud tasks, IDEs, GitHub, Slack, and Linear. Here is how to use it and how it compares to Claude Code.
5 min readGPT-5 Codex: OpenAI's Agentic Coding Model
OpenAI is drawing a line in the sand. GPT-5 Codex is not an API release.
7 min readCodex Is Becoming a General-Purpose AI Agent, Not Just a Coding Tool
OpenAI is turning Codex from a coding assistant into a broader agent workspace for files, apps, browser QA, images, automations, and repeatable knowledge work.
8 min readNew here? Start with
Technical content at the intersection of AI and development. Building with AI agents, Claude Code, and modern dev tools - then showing you exactly how it works.








