GPT-6 Astra is here: OpenAI’s new AI model can use computers, browse web and get work done

OpenAI has launched GPT-6 Astra, its latest frontier AI model, as the company pushes AI further from simply generating answers to actually getting work done.

OpenAI CEO Sam Altman said the company hopes Astra will enable “a new generation of entrepreneurship, scientific discovery, and building”, calling it the world’s best model for computer use, professional work, science, coding and cybersecurity.

Altman also said the launch took extra time because OpenAI wanted to ensure Astra met the safety and alignment standards required for a model with these capabilities. “We think you’ll find it worth the wait,” he said in a post on X.


Astra is being positioned as a model that can take a goal and work through multiple steps to complete it.

What is GPT-6 Astra?

GPT-6 Astra is OpenAI’s newest AI model, built on advances in pre-training, reinforcement learning and alignment.

The company says Astra is its most intelligent and aligned model yet. It is designed to use computers, browse the web, write and test software, analyse scientific data and handle complex professional tasks with less human intervention.

Also Read: How OpenClaw 2.0 is trying to make AI agents easier to use

For instance, it can fill online forms, update customer records, organise calendars, conduct online research and draft summaries in email or document editors. It can also analyse scientific data, create plots, build websites and test whether those websites actually work.

Why professional work is a major focus

OpenAI says that the model can create documents, presentations, spreadsheets and analyses while following existing templates and matching a company’s writing or visual style.

It is also said to be designed to make better decisions when instructions are incomplete.

If a missing piece of information could materially change the outcome, Astra can ask a question. If the decision is routine, it can make a reasonable assumption and continue.

The company also adds that Astra is better at handling changes during a task without losing sight of the original objective.

Another important change comes to Codex. Astra can retain notes across context windows and search earlier context when needed, rather than relying entirely on a compressed summary of a long session. This could be particularly useful for large coding projects where important details can get lost as conversations grow.

The cybersecurity boost comes with risks

OpenAI says the model represents a significant jump in cyber capabilities and meets the “Critical” threshold under its Preparedness Framework.

According to the company, Astra discovered and used two previously unknown zero-day vulnerabilities during testing. The company says it has disclosed both vulnerabilities to their maintainers.

Expert assessments are also said to have found that Astra could use previously unknown vulnerabilities to achieve arbitrary code execution in hardened browsers and develop privilege-escalation exploits for hardened operating systems.

That capability could help security teams identify and fix vulnerabilities faster. But the same capabilities could also make it easier for attackers to exploit software.

As a result, the version being launched today will refuse more advanced cybersecurity requests, including creating proof-of-concept exploits. OpenAI says it plans to expand access to defensive security workflows through its Daybreak programme.

What has OpenAI done on AI safety?

Safety and alignment are a major part of the Astra launch, particularly given the model’s ability to take actions autonomously.

OpenAI says that the company has also built monitoring systems that watch the model’s reasoning and actions for potentially unauthorised behaviour and can stop activity when necessary.

There is, however, a caveat. OpenAI says Astra’s written reasoning was harder to monitor than GPT-5.6 Sol’s in tests designed to make the model evade monitoring. The company says improving this “monitorability” remains a research priority.

Also Read: OpenAI is building 'automated shutdown' capabilities for AI tools, letter to lawmakers says

OpenAI also says Astra was better at staying within authorised boundaries in an evaluation inspired by the Hugging Face incident. Without production safeguards, GPT-5.6 Sol went beyond the authorised target in 48% of cases on an “impossible” task evaluation, while Astra did so in 0% of cases.

When can you use GPT-6 Astra?

Astra is rolling out initially to a limited set of organisations. Over the coming days, it will become available to ChatGPT Plus, Pro, Business and Enterprise users, as well as through the OpenAI API and AWS.

Developers will be able to access it as gpt-6-astra through the OpenAI API and Amazon Bedrock.

OpenAI says API pricing starts at $10 per million input tokens and $50 per million output tokens. A faster mode, which can deliver up to twice the speed, costs twice the standard price.

Thibault Sottiaux, Head of Core Products at OpenAI, said in a post on X that the company is also offering paid ChatGPT users a form of compensation for the staggered rollout. Users will receive one “banked reset” for every day they remain without access to GPT-6 Astra.