If ChatGPT Astra is missing from your account, you are not doing anything wrong. OpenAI released GPT-6 Astra on 3 September 2026 and is rolling it out in stages, starting with a limited set of organisations before reaching all Plus, Pro, Business and Enterprise users.
This page covers who has access today, how to switch models, and what to do while you wait. For what the model does, read our GPT-6 Astra explained guide.
Rollout status as of 5 September 2026: Astra is live in ChatGPT for a growing share of paid accounts, live in Codex, and live for developers via the API, Azure and AWS Bedrock.
Business and Enterprise workspaces still need an admin to turn it on, and Free-tier availability remains unconfirmed. We update this page as the rollout progresses.
Who can use ChatGPT Astra right now?
As of 5 September 2026, ChatGPT Astra is reaching paying ChatGPT users in stages. OpenAI began with a limited set of organisations under a programme it calls Daybreak Access, then widened to all Plus, Pro, Business and Enterprise accounts “over the coming days”. Developers can also use it through the API, Azure and AWS Bedrock.
“Over the coming days” is OpenAI’s own language, so two people on identical Plus subscriptions can have different experiences on the same morning. The OpenAI announcement post sets out the sequence, and the rollout has visibly confused Plus subscribers. There is no queue to jump: Daybreak Access was an OpenAI-selected programme for organisations, not an open waitlist.
Which ChatGPT plans include GPT-6 Astra?
Every paid ChatGPT plan is in scope: Plus, Pro, Business and Enterprise. Pro gets the largest included allowance, Business and Enterprise need an admin to enable the model first, and Free-tier access has not been confirmed by OpenAI at all.
| Plan | Astra access as of 5 Sept 2026 | Who switches it on | Notes |
|---|---|---|---|
| Free | Not confirmed by OpenAI | Not applicable | Assume you do not have it until OpenAI says otherwise. |
| Plus | Rolling out | Automatic, no action needed | Appears in the picker when your account is reached. |
| Pro | Rolling out, largest allowance | Automatic, no action needed | Most usage headroom of the consumer plans. |
| Business | Rolling out, off by default | Workspace admin or owner | Nobody sees Astra until an admin enables it. |
| Enterprise | Rolling out, off by default | Workspace admin or owner | Same as Business, often after an internal review. |
| API / Azure / Bedrock | Available | Your account or cloud admin | Pay per token. Tool calling needs the Responses API. |
Is ChatGPT Astra free on the free plan?
OpenAI has not confirmed Free-tier access to GPT-6 Astra. Its rollout statement named Plus, Pro, Business and Enterprise, and said nothing about the free plan. Any article telling you Astra is free today is guessing, and on the free plan you should not expect it yet.
Astra is also expensive to run: $10 per million input tokens and $50 per million output, 2.5 times what GPT-5.6 Sol cost. Models at that price reach free plans slowly, in limited form, or not at all. If you are staying on the free plan for now, our roundup of free AI tools covers what you can use today without waiting for a rollout.
How do I switch to Astra in ChatGPT on web, desktop and mobile?
Once Astra reaches your account, you pick it from the model picker at the top of a chat, exactly as you would any other model. The steps are near-identical on web, desktop and mobile, and the choice applies to the conversation you are in, not your whole account.
On the web (chatgpt.com):
Step 1: Start a new chat.
Sign in to your account and open a fresh conversation.
Step 2: Open the model picker.
Click the model name at the top of the conversation.
Step 3: Choose GPT-6 Astra.
If there is a “more models” submenu, check inside it.
Step 4: Send your message.
The model name stays at the top so you can confirm which model answered.
On the desktop app (Windows and macOS): fully quit and reopen the app so it picks up your latest account state, open the model picker at the top of a new chat, then select GPT-6 Astra.
On mobile (iOS and Android): update the ChatGPT app first, because model rollouts often need a current build. Then open a new chat, tap the model name at the top of the screen and pick GPT-6 Astra from the sheet.
Set the model before your first message. Switching mid-conversation works, but a fresh thread gives a cleaner comparison when judging output quality.
Why is Astra not showing in my model picker yet?
The most common reason is that the staged rollout has not reached your account, and there is nothing to fix. The other causes are an outdated app build, a cached browser session, a Business or Enterprise workspace where an admin has not enabled the model, or a regional restriction. Work through them in that order.
Work the checklist:
- Check your plan. Astra is going to Plus, Pro, Business and Enterprise. On the free plan, OpenAI has not confirmed access.
- Check which account you are in. Personal and work accounts have different model pickers, and it is easy to be looking at the wrong one.
- Update and restart. Update the mobile app, reopen the desktop app, or hard-refresh the web page.
- Ask your admin. On Business and Enterprise, Astra is off by default, and many organisations run a security review before enabling it.
- Rule out an outage. Our guide to checking OpenAI status when ChatGPT is down covers where to look, including the 3 September 2026 outage that hit ChatGPT and Codex.
Signing in and out of a personal account and a work account all day is a common source of this confusion, and our guide to AI tools for remote work covers how distributed teams keep those two setups straight.
If all five check out, waiting is the correct action. Rollouts this size finish over days, not hours.
Build A Baseline While You Wait
Use the wait to run three real tasks from your own week through your current model and save the outputs verbatim. When Astra lands you can rerun the same three and judge the difference on your own work instead of on a demo prompt.
How does a Business or Enterprise admin turn Astra on?
Astra is off by default in Business and Enterprise workspaces, so an owner or admin must enable it before any member can use it. The control sits in workspace admin settings, under the model or feature controls where new models are switched on per workspace. Astra then appears in members’ pickers on their next session.
Settle two things first. Astra reached OpenAI’s “critical” cybersecurity capability threshold, meaning it can find previously unknown vulnerabilities without human guidance, and OpenAI ships it with safeguards restricting it to defensive use. Your security team will want to have read that rather than discovered it.
Cost is the other. Independent testing puts Astra at roughly 75% more expensive per task than its predecessor, despite using about 10% fewer output tokens. Our breakdown of the GPT-6 Astra benchmarks shows where that buys real improvement and where the model regressed.
Enable Astra For A Pilot Group First
Switch Astra on for one small team before the whole workspace, and record your current cost per task that same week so finance has a before number. A 75% per-task rise is far easier to approve when you can show which tasks it actually improved.
How do I use ChatGPT Astra in Codex?
Astra is available in Codex, OpenAI’s coding agent, alongside the ChatGPT rollout. Select it as your Codex session model the same way you would in ChatGPT. The headline change for Codex users is asynchronous questioning, which lets the model raise a question without halting the run.
It also holds context across long sessions with searchable notes, and OpenAI says it respects task boundaries better than the previous generation. The trade-off is that it asks for clarification more readily, which can stall autonomous work. Our guide on how to prompt GPT-6 Astra covers the instructions that counter that.
How do I get GPT-6 Astra API access through OpenAI, Azure and AWS Bedrock?
Astra is available through the OpenAI API, Microsoft Azure and AWS Bedrock. With OpenAI directly you call it from an existing API account, no separate application needed. On Azure and AWS Bedrock you enable model access in your cloud console, which usually means an account admin rather than an individual developer.
Two technical points will save you an hour of debugging. Tool calling requires the Responses API, not Chat Completions: that is where web search, file search, image generation, code interpreter, computer use, hosted shell, MCP, skills and tool search live. And temperature, top_p and top_logprobs are no longer supported, while prompt_cache_retention is replaced by prompt_cache_options.ttl set to 30m.
Instead of a temperature dial you use reasoning.effort, with five levels: low, medium, high, xhigh and max. The none setting from earlier models does not exist here, so migrate from none or minimal by starting at low. Astra takes 1,050,000 input tokens, returns up to 128,000, accepts text and images, and has a knowledge cutoff of 30 April 2026. Full details are in the GPT-6 Astra model reference.
What are the ChatGPT Astra usage limits, and what does Fast mode cost?
ChatGPT subscribers get an Astra allowance included with their plan and can buy additional credits. Limits vary by plan, with Pro getting the most headroom. On the API you pay per token: $10 per million input, $50 per million output, $1 per million for cached input.
Fast mode runs at twice the speed for twice the price, so $20 and $100 per million tokens. It is not available for EU data residency, a real constraint if your organisation has committed to keeping data in the EU. Batch pricing is there if latency does not matter.
GPT-5.6 Sol cost $4 and $20 per million tokens, making Astra a 2.5x increase per token before you count how many tokens a task consumes. Budget for it. If words like tokens, allowances and cached input are new to you, our plain English LMS and learning technology terminology glossary defines the vocabulary vendors use around AI features.
Conclusion
If Astra has reached your account, spend your first session comparing it against yesterday’s model on a real task rather than a demo prompt, because the independent numbers show gains in some areas and regressions in others. If it has not, ask your admin whether it is enabled, then check back in a day or two.
Either way, read the benchmark detail before committing budget. Astra is better at some things and more expensive at all of them, and knowing which is which is the difference between a good decision and a large invoice.
FAQ
Q1. Is GPT-6 Astra free to use?
OpenAI has not confirmed Free-tier access to Astra. The rollout named Plus, Pro, Business and Enterprise plans plus API access. On a free ChatGPT account, assume you do not have it. Paid plans include an Astra allowance, and API users pay $10 per million input tokens and $50 per million output.
Q2. How long will the ChatGPT Astra rollout take?
OpenAI said Astra would reach all Plus, Pro, Business and Enterprise users “over the coming days” after the 3 September 2026 release, without naming a completion date. As of 5 September 2026 it is still in progress, and there is no way to accelerate it on a personal plan.
Q3. What is Daybreak Access?
Daybreak Access is the name OpenAI used for the first stage of the Astra rollout, which went to a limited set of organisations before the wider release to paid ChatGPT plans. It was not an open waitlist individuals could join, so there was no sign-up form you missed.
Q4. Why can my colleague use Astra when I cannot?
Two likely reasons. Either the staged rollout has reached their account and not yours, which resolves on its own, or you are in a Business or Enterprise workspace where Astra has not been enabled while they use a personal Plus account. Check which account your session is using.
Q5. Do I need the Responses API to use Astra?
You need it for tool calling. Plain text generation works through the standard interface, but every hosted tool, including web search, file search, code interpreter, computer use, MCP and skills, requires the Responses API. Chat Completions does not support tool calling on Astra at all.
