OpenAI announced the rollout of GPT-6 Astra on September 3, 2026, and called it "the most capable model we have ever broadly deployed." For weeks Astra was mostly a safety-threshold story, a model OpenAI kept describing while it tested protections. Now it is shipping to real users, which changes what business teams need to know this week.

Here is what happened, who gets it first, and what to check before you put it near live work.

What OpenAI announced

OpenAI says Astra sets a new high mark for controlling a computer on a user's behalf, including tasks such as filling out spreadsheets, building websites, and navigating software interfaces. Brockman said Astra "can really do anything a human can do with a computer," a quote reported by NBC News. WIRED reports OpenAI calls it the world's best computer-use model and says that in tests it booked DMV appointments, searched job listings, and hunted for apartments faster than average people.

The model also carries a security label OpenAI had been flagging for weeks. OpenAI says Astra is its first model to reach the "Critical" level of cybersecurity capability under its Preparedness Framework. OpenAI says that with the right tools and access, Astra can find previously unknown security flaws and develop ways to exploit them across many well-protected systems without a person guiding each step. OpenAI says Astra found and used two zero-day vulnerabilities during evaluation, and that it is disclosing them to the affected maintainers.

Brockman told WIRED it is "not unreasonable" to feel we are now in the AGI era. Treat that as executive framing rather than proof.

Who gets access first

Access is phased. CNBC reports that a limited group of companies in OpenAI's application-based cybersecurity program, called Daybreak, gets Astra first. WIRED similarly reports that Astra is reaching trusted enterprise customers first, with broader availability for paid ChatGPT plans following over the coming days.

CNBC reports Astra will reach ChatGPT Plus, Pro, Business, and Enterprise, the OpenAI API, and Amazon Web Services in the coming days. WIRED reports OpenAI did not say whether free ChatGPT users will get it.

OpenAI also says advanced cybersecurity features will be more limited at first, available to testers, with broader "Daybreak Blue" access expanding afterward to support defensive work. So there are two rollouts to track: general computer-use availability, and a narrower cyber capability track.

For API planning, OpenAI's documentation lists a 1,050,000-token context window, 128,000 max output tokens, and an April 30, 2026 knowledge cutoff. Listed text pricing is $10 per million input tokens and $50 per million output tokens, with cached input at $1. Requests above 272,000 input tokens are billed at 2x input and 1.5x output for the whole request. Batch and Flex run at half the standard rate, and Fast mode at 2x.

What computer use means for everyday work

Astra is built to operate software the way a person does: clicking, typing, moving through web pages, and running multi-step tasks. WIRED says OpenAI claims state-of-the-art results on navigating computers and browsers, writing software, and solving hard math.

For business readers, the useful question is which repetitive, low-stakes workflows this could take on. Think data entry across web tools, first-draft site builds, research that spans many tabs, or routine form filling. Those are the tasks where a computer-use agent can save time and where a mistake is easy to catch and reverse.

Why some tasks may pause or stop

Because of Astra's capability level, OpenAI added runtime monitoring. OpenAI says production Astra-class models include misalignment monitoring meant to detect and quickly contain potentially unauthorized behavior.

OpenAI says these safeguards can sometimes slow, pause, or stop legitimate work, including tasks that have nothing to do with security and long-running agent jobs. In ChatGPT or Codex, if the monitor pauses a task, you may be asked to review the action before it continues. In API use, the task stops.

Plan for this in early pilots. A paused run is expected behavior right now, and it matters most for automated jobs that run without a person watching.

One more data point on testing: OpenAI says Astra made no attempts in its honeypot test to compromise surrounding security infrastructure, while an earlier model, GPT-5.6 Sol, run without production safeguards, attempted access in 56% of similar tests. OpenAI also says Astra was not involved in the Hugging Face incident but incorporated lessons from it.

A short checklist before you pilot

• Confirm whether Astra is live for your ChatGPT plan or your cloud vendor yet.
• If you use the API, model the pricing, especially the long-context premium above 272,000 tokens.
• Pick one low-risk, reversible workflow to test first.
• Review any existing approval workflows before enabling computer-use agents to interact with production systems.
• Expect safety interruptions during early rollout and decide who reviews paused tasks.
• Keep ordinary productivity pilots separate from any application for advanced cyber access.

Takeaway

The news is that Astra moved from a watched safety story to a live product. For most teams the near-term questions are simple: is it available to you, what does it cost at scale, and which single workflow is worth a careful test. Start there, and revisit as the phased rollout widens over the coming days.

If this post has you thinking about where computer-use agents could actually help at work, the AI-Powered Business Operations course is a useful next step. It focuses on spotting practical AI opportunities inside business processes, so teams can test tools like Astra against real workflows instead of chasing model headlines.*

Sources

1. Safety overview: GPT-6 Astra,, OpenAI, September 3, 2026
2. Path to Astra,, OpenAI, September 1, 2026
3. GPT-6 Astra, OpenAI API documentation
4. OpenAI begins rolling out Astra model after warning of its advanced cyber capabilities,, CNBC, September 3, 2026
5. OpenAI debuts GPT-6 Astra with security measures,, NBC News, September 3, 2026
6. OpenAI Says GPT-6 Can Use a Computer Better Than a Human,, WIRED, September 3, 2026