OpenAI released Astra on Thursday, September 3, 2026, describing it as its most powerful model to date, according to TechCrunch and Wired. The model is being made available first to customers in OpenAI's Daybreak cybersecurity early-access program, then over the following week through Pro, Plus, Enterprise, and Business plans and the API. OpenAI did not say whether it would reach free ChatGPT users, Wired reported. OpenAI president Greg Brockman called Astra the company's "most intelligent and, also very importantly, our most aligned model yet." OpenAI claims it is the "world's best computer use model" and the "best model for software engineering to date," citing benchmark results it says exceed its own Sol model and Anthropic's Fable. In tests, OpenAI said Astra booked DMV appointments, searched job listings, and apartment-hunted faster than an average person. TechCrunch reported that Astra uses a technique called opaque recurrence, which can obscure chain-of-thought monitoring used to audit model decisions. Chief scientist Jakub Pachocki said that "as model capabilities are increasing, monitorability is getting more challenging," adding OpenAI would "withhold scaling" if it could not maintain confidence in monitoring. Brockman said he personally believes the world is now in the AGI era, noting a prior contractual AGI clause with Microsoft no longer exists.
- Released September 3, 2026, first to Daybreak cybersecurity program, then paid plans and API
- OpenAI claims it is the best model for computer use and software engineering, citing benchmarks against Sol and Fable
- Uses 'opaque recurrence,' which can obscure chain-of-thought monitoring, per TechCrunch
- Brockman says he personally believes the AGI era has arrived; prior Microsoft AGI contract clause no longer exists
What it means for you
OpenAI has a new flagship model aimed at doing computer tasks for you — clicking through websites, filling forms, writing code. The headline claims are about speed and skill, but the more notable detail is that the model's internal reasoning is harder to inspect, and OpenAI itself says keeping tabs on advanced models is getting difficult. For most people this is a preview of where the tools are heading, not something you need to act on today.
Try this
If you already pay for ChatGPT, watch for Astra to appear in your plan over the coming days and test it on one repetitive web task you'd normally do by hand — then check its work rather than trusting it blind.
Who should care
Developers, security teams, and anyone building on OpenAI's API or planning to hand off web-based tasks to an AI agent.
Skip this if
You use AI mainly for writing, summarizing, or answering questions — the practical difference from your current model will be marginal, and the AGI talk changes nothing about your day.
Sources: TechCrunch AI, Wired AI — read the original