⚠️ Affiliate Disclosure: This post is not an affiliate associated review. It may become one in the future and this disclosure will reflect that. This does not affect my reviews and I only recommend tools I genuinely believe in.
OpenAI has announced GPT-6 Astra, the new flagship model behind ChatGPT — and the headline for most of our readers is who gets it. Astra is rolling out over the coming days to Plus, Pro, Business and Enterprise subscribers at no extra cost, with a Pro-grade version for higher tiers. The free plan is not mentioned in the rollout at all. Here is what Astra actually does, what OpenAI is claiming, and what it means if you use ChatGPT — our 9/10-rated pick for most people.
What Is GPT-6 Astra?
Astra is OpenAI’s successor to the GPT-5.6 generation, pitched as “the world’s most intelligent and aligned model”. Strip the superlatives and three concrete claims stand out: it is markedly better at using a computer (filling forms, updating CRMs, researching and drafting in your actual documents), better at finished professional work (slides, spreadsheets and documents that follow your templates), and better at knowing when to ask — OpenAI says it asks focused questions when missing information would change the outcome, rather than guessing.
What OpenAI Is Claiming
- State-of-the-art computer use: on OpenAI’s own OSWorld 2.0 figures, Astra completes tasks at higher accuracy in roughly half the time of its predecessor
- Best coding model to date, by their benchmarks — including numbers published head-to-head against Anthropic’s Claude Fable 5.1
- Real mathematics: OpenAI says Astra helped tighten a prime-gap bound that had stood for over a decade, and improved a term unchanged for 80 years
- A first: Astra meets the Critical threshold for cybersecurity under OpenAI’s own Preparedness Framework — it found genuine zero-day vulnerabilities during testing, so it ships with restrictions on advanced security tasks
All of those are OpenAI’s figures from OpenAI’s launch page — vendor benchmarks are marketing until independently tested, and the launch table’s footnotes sniping at rivals’ grading methods rather prove the point. We have not yet tested Astra hands-on; it reaches Plus accounts over the coming days, and we will run it through real work when it lands.
Who Gets GPT-6 Astra — and Who Doesn’t
Plus, Pro, Business and Enterprise users get Astra within their existing subscription allowances — no price rise announced — and can buy extra usage credits. Pro and business tiers also get GPT-6 Astra Pro. Enterprise admins must switch it on; it is off by default. And the notable omission: the free plan. Free ChatGPT stays on the current models for now, which makes this the clearest capability gap OpenAI has drawn between its free and paid tiers in some time. Developers get it as gpt-6-astra at $10 per million input tokens and $50 per million output.
The Benchmark War With Claude
OpenAI’s launch page publishes direct comparisons against Anthropic’s Claude Fable 5.1, released just days earlier — and claims wins on most rows it chose to print. Two things can be true: Astra’s numbers are genuinely strong, and a vendor’s launch table shows the benchmarks the vendor picked. Our own testing has Claude ahead on long-form writing and ChatGPT ahead as the all-rounder; whether Astra changes that is exactly what we will test.
Should You Care?
If you pay for ChatGPT Plus: yes — you are getting the new flagship at no extra cost within days, and the computer-use and document claims target exactly the everyday work our readers do. If you use free ChatGPT: nothing changes yet, and that is the story — the free tier that was “far more capable than most people realise” in our review now sits a full model generation behind the £16/month one. If that gap holds in real use, it changes our value-for-money calculus, and we will say so in the updated ChatGPT review.
Frequently Asked Questions
Is GPT-6 Astra free to use?
No. Astra is rolling out to paid ChatGPT tiers — Plus, Pro, Business and Enterprise — included in existing subscriptions. OpenAI’s announcement does not include the free plan, which stays on current models.
What is GPT-6 Astra best at?
By OpenAI’s account: using a computer like a person (forms, CRMs, research, building and testing websites), producing finished documents, slides and spreadsheets that follow your templates, and coding. These are vendor claims we have not yet verified hands-on.
Is GPT-6 Astra better than Claude?
OpenAI’s own launch benchmarks say yes on most published rows against Claude Fable 5.1 — but vendors choose which benchmarks to print. In our independent testing, Claude leads on long-form writing and ChatGPT on all-round versatility; both scored 9/10. We will retest both on real work now the new models are out.
Source: OpenAI’s GPT-6 Astra announcement, September 2026. We will update this page after hands-on testing.
Leave a Reply