On September 3, 2026, OpenAI unveiled its new flagship model, GPT-6 Astra. It is rolling out over several days through paid ChatGPT plans, the API (application programming interface — the connection that lets developers call an AI model directly from their own software), and AWS (Amazon Web Services — Amazon's cloud computing platform, also used as a channel for offering AI models from various companies to businesses). This doesn't change Claude's plans or features, but since it's a topic Claude users have been asking about, here's what was announced and what it means for you.
🟢 Model references match the current lineup · model notice · Fable subscription
Fable 5 and 5.1 subscription (updated November 1, 2026): Claude Fable 5.1, released September 1, 2026, is the current Fable model and Fable 5 is now legacy. Plan terms are the same for both — Max, Ultra, and Team Premium plans include Fable at up to 50% of the weekly usage limit; Pro and Team Standard use usage credits ($10/M input, $50/M output tokens). The one-time $100 credit applied only to the Fable 5 transition and is not offered for 5.1. Some coding and debugging requests may be answered by an Opus model due to a security classifier (both models). See the Fable 5.1 guide and the Fable 5 availability guide for details.
What is GPT-6 Astra?
GPT-6 Astra is OpenAI's new flagship AI model, unveiled in September 2026 as the successor to GPT-5.6 (released July 2026). OpenAI described it as a "generational leap," and OpenAI president Greg Brockman told reporters it could eventually be seen as the start of AGI (Artificial General Intelligence — a term for a hypothetical level of AI that can think and learn broadly across many domains the way a person can, rather than being limited to specific tasks). That said, this is a comment from a company executive; OpenAI has not officially declared that AGI has been achieved.
What can GPT-6 Astra actually do?
The most notable change is the speed of its computer use (computer use — a feature that lets an AI look at a screen and operate it like a person would, moving a cursor and typing) capability. According to OpenAI, GPT-6 Astra is about twice as fast at computer-use tasks as its predecessor, handling things like organizing spreadsheets, filling out forms, and navigating web pages by looking at the screen the way a person would.
This concept itself isn't new to the industry. Anthropic has offered a computer use tool in the developer-facing Claude API for some time, and recently added a browser use tool that lets Claude access the web directly. This announcement is best understood as OpenAI extending the same concept to its consumer ChatGPT product with a significant speed boost.
Why is it being called a "risky model"?
OpenAI has officially stated that GPT-6 Astra is the first model to cross the "Critical" cybersecurity capability threshold under its Preparedness Framework (Preparedness Framework — OpenAI's own safety evaluation system for checking a model's risk level before it gains dangerous capabilities). This means that, under the right conditions, the model could find previously unknown security vulnerabilities and develop ways to exploit them without human guidance.
Because of this, the version deployed to general paid users is designed to automatically refuse advanced offensive requests, such as writing exploit code. Access with relaxed restrictions is limited to organizations vetted through Daybreak (Daybreak — an OpenAI program that grants lower-restriction access to organizations verified to have a legitimate purpose, such as cybersecurity research).
- General paid users — use a restricted version that automatically refuses advanced attack-related requests.
- Vetted Daybreak organizations — can access a version with relaxed restrictions if approved for a legitimate purpose such as cybersecurity research.
Note: This article does not repeat any self-reported benchmark scores or claims that the model "beats" a competitor. Such figures vary widely depending on how they're measured, and repeating them uncritically can be misleading.
How much does it cost, and how does it compare to Claude?
GPT-6 Astra's API pricing is $10 per million input tokens and $50 per million output tokens. Coincidentally, this is exactly the same rate as Anthropic's specialized Claude Fable 5.1 model. Anthropic's general-purpose flagship, Claude Opus 5.5 (released September 22, 2026), is cheaper at $4 per million input tokens and $20 per million output tokens (the previous Opus 5 was $5/$25). The table below summarizes this.
| Model | Provider | Type | Input (per million tokens) | Output (per million tokens) |
|---|---|---|---|---|
| GPT-6 Astra | OpenAI | General-purpose flagship | $10 | $50 |
| Claude Opus 5.5 | Anthropic | General-purpose flagship | $4 | $20 |
| Claude Fable 5.1 | Anthropic | Specialized model | $10 | $50 |
On the ChatGPT side, it's rolling out to Plus ($20/month), Pro ($100/month), Business ($200/month), and Enterprise users. Whether it's available on the free plan has not been confirmed in official announcements.
Does this news matter for Claude users?
Nothing about Claude's plans or features has changed as a result of this announcement. As of this writing (September 8, 2026), no official response from Anthropic has been confirmed. GPT-6 Astra is simply a new model OpenAI added to its own product lineup, and there's no indication yet that Claude users need to change their current plan or workflow. That said, if you're interested in computer-use style features, it's worth knowing Anthropic already offers something similar through its developer API.
Frequently Asked Questions
Q. Can I use GPT-6 Astra for free?
Based on what's been confirmed, it's being rolled out through ChatGPT Plus and higher paid plans, plus the API and AWS. Availability on the free plan has not been confirmed in official announcements.
Q. Is GPT-6 Astra really AGI?
OpenAI has not officially declared that AGI has been achieved. President Greg Brockman suggested it could be seen as the start of AGI in media comments, but the industry itself doesn't agree on a single definition or threshold for reaching AGI.
Q. Does Claude have a computer-use feature too?
Yes, Anthropic has offered a computer use tool in the developer-facing Claude API for some time, and has more recently added a browser use tool as well.
Q. Why does GPT-6 Astra refuse certain requests?
Because OpenAI classified it as reaching a "Critical" level of cybersecurity capability under its own framework. As a result, the version given to general paid users is designed to automatically refuse advanced offensive requests such as writing exploit code.
Related articles
- Why Claude Suddenly Switches to a Different Model Mid-Conversation
- Claude vs ChatGPT: Key Differences and When to Use Each (2026)
- Claude vs Gemini: Key Differences and Which to Choose
- Claude Models Compared: Haiku vs Sonnet vs Opus, and Which to Use
- Claude Opus 4.8: The Complete Guide to What's New and How to Use It
- What Is the GPT-6 Astra 'Sweet Spot'? Choosing a Reasoning Effort Setting (and How Claude Compares)