How to Use ChatGPT Astra 6 (GPT-6 Astra) in Work, Codex and API

To use “ChatGPT Astra 6” — officially GPT-6 Astra — you open ChatGPT Work or Codex on a paid plan (Plus and up) and pick GPT-6 Astra there, or, on Pro, Business and Enterprise, use GPT-6 Pro in regular Chat, which OpenAI says is powered by Astra. Developers can call it directly in the API with the model ID gpt-6-astra.

That’s the short version. The confusing part, and probably why so many people are searching right now, is that Astra does not simply show up as a new option in the normal chat box for everyone. Below is where to find it on each plan, step by step, plus a few tips for getting good results. Details are as of October 2026 and OpenAI has been adjusting the rollout, so check the linked Help Center pages if something looks different.

What GPT-6 Astra is

OpenAI announced GPT-6 Astra on September 3, 2026, as its most capable model so far. It is built for long, multi-step work rather than quick back-and-forth chat: navigating websites and software, keeping track of a goal, and continuing to work through a problem, as Tom’s Guide describes it. Tom’s Guide calls it the successor to GPT-5.6 Sol.

“ChatGPT 6” is how many people describe it, but OpenAI doesn’t use that as a separate brand. The model is GPT-6 Astra, and you reach it through ChatGPT’s surfaces or the API.

Access started with a limited set of organizations before reaching paid ChatGPT plans and the API. Under OpenAI’s Preparedness Framework it is also the first OpenAI model rated Critical for cybersecurity capability, which is one reason the rollout began with vetted groups.

Which plans can use Astra, and where

This is the bit that trips people up. According to OpenAI’s GPT-6 Pro Help Center article and Notebookcheck’s reporting:

A few caveats:

• OpenAI’s announcement mentioned Plus users, and that led to a community thread asking for clarification. As of this writing the Help Center says Plus access is limited to Work and Codex.
• On Enterprise, your workspace admin’s model-access settings also decide whether you see it.
• Some third-party guides claim Plus users can pick Astra from the regular model dropdown. That contradicts OpenAI’s own Help Center, so trust the Help Center.

How to use GPT-6 Astra in ChatGPT Work

ChatGPT Work is OpenAI’s agent mode for longer, multi-step tasks that end in a finished deliverable, such as research, analysis, documents, spreadsheets or presentations. OpenAI’s Work and Codex article describes how to get there:

1. Update the ChatGPT desktop app to the latest version. If Astra is missing, OpenAI suggests checking for updates through the app’s menu even if you updated recently.
2. Click ChatGPT in the top-left menu.
3. Use the toggle at the top to switch from Chat to Work.
4. Choose GPT-6 Astra in the model and effort control. Reports describe the default setting as a single slider that combines model and reasoning level.
5. Describe the task and the result you want, for example “Compare these three vendors and produce a one-page summary table.”

Note that ChatGPT Work was rolling out gradually to eligible accounts, so it may not appear on every account at the same time.

Tips for Work

• Describe the finished deliverable, not just the question. Astra is built to plan and carry out several steps, so tell it what the end product should look like.
• Expect pauses. OpenAI added extra monitoring for cases where an agent may have misread your instructions, and a task can pause so you can review it. Read what it’s asking before you tell it to continue.
• Voice works here too. ChatGPT Voice can drive Work on web, mobile and the desktop app.

How to use GPT-6 Astra in Codex

Codex is OpenAI’s software-development agent, and it’s the other place Plus users get Astra.

In the ChatGPT app

1. Click ChatGPT in the top-left corner and choose Codex from the dropdown.
2. Click the effort control at the bottom-right of the message box.
3. Select GPT-6 Astra and an effort level.
4. Describe the coding task as you normally would.

In the Codex CLI

1. Update the CLI. OpenAI says Astra requires Codex CLI 0.153.0 or newer.
2. Start it with the model flag: codex -m gpt-6-astra.
3. If you set effort in your Codex config, note that the “Light” setting maps to low for Astra.

One community forum post reported the Codex Windows app showing “GPT-6 Astra” while actually requesting a different model. That’s a single unverified report, but if results seem off, it’s worth checking which model your requests actually use.

Usage limits and effort levels

On Plus, Astra usage in Work and Codex draws from your plan’s included Work and Codex allowance, whether you run it at Standard speed or in Fast mode. Once that’s used up, OpenAI lets you buy more credits if you want to keep going. Exact limits differ by plan and have changed during the rollout, so see OpenAI’s usage management article for current numbers.

Effort is the setting that matters most for your allowance. OpenAI’s guidance is that lower effort on Astra can match or beat higher effort on the older Sol model, so start at Low or Medium and only go up when a task clearly needs it.

How to use GPT-6 Astra in the API

Developers don’t need a ChatGPT plan. API access is billed separately. According to the GPT-6 Astra model page and OpenAI’s developer guides:

• Model ID: gpt-6-astra.
• Reasoning effort: low, medium, high, xhigh and max. Unlike some earlier models, it does not accept none. Sending it returns an HTTP 400 error, and OpenAI’s migration guide says to use low instead.
• Use the Responses API for tools. Function calling with this model is supported in the Responses API, not in Chat Completions.
• Changing effort mid-conversation: with Responses, you can send a configuration_update input item, which keeps the original prompt prefix so caching still works.
• Speed tiers: the docs describe Fast and Ultrafast modes, and note that Fast mode for Astra doesn’t come with a latency SLA.

Pricing and context-window figures are published on OpenAI’s model page. Check there rather than relying on numbers repeated in blog posts, since several widely shared figures did not come from OpenAI.

Troubleshooting: “I can’t see Astra”

1. Check your plan. On Free or Go, Astra isn’t included. On Plus, look in Work or Codex, not regular Chat.
2. Update the app or CLI. Older desktop app versions and Codex CLI versions below 0.153.0 won’t show it.
3. Ask your admin if you’re on Business or Enterprise. Workspace permissions can hide models.
4. Give it time. Work and some Astra features have been rolling out gradually, so availability can differ between accounts for a while.

If you’ve checked all four and it’s still missing, OpenAI’s Help Center and support are the right places to go. Don’t use third-party “unlock” tools or share your login with anyone who claims they can enable it for you.