OpenAI launches GPT-6 Astra, its strongest mannequin but and touts its capacity to make use of your pc

It’s just one of a wide range of new and advanced capabilities, OpenAI says, including mathematical proficiency, software engineering, and defensive cybersecurity. But “computer use is a particularly important part of what’s new,” OpenAI co-founder and president Greg Brockman told reporters today. The model “can zip through spreadsheets, fill out forms, and navigate across webpages often at superhuman speed,” he said.

The company shared a video with reporters of someone sitting on a chair, instructing the computer through their voice in what appeared to be a seamless, albeit staged, interaction.

Computer use “shows how far we’ve come from sort of aspirationally training for computer use to bringing real value to people every day,” Mia Glease, vice president of research at OpenAI, said.

OpenAI also released benchmark assessments that showed Astra beats its previous best model, GPT-5.6 Sol, as well as the best current models from its rival Anthropic, Claude Fable 5.1, on a wide range of tasks. On ARC-AGI-3, a difficult benchmark that is meant to test a model’s ability to apply reasoning to situations it has not encountered before, Astra scored 98.6%, whereas GPT-5.6 Sol had only scored 7.8% and Anthropic’s Claude Opus 5 had only scored 30%.

Astra also scored 100% on ExploitGym, a tough cybersecurity challenge beating GPT-5.6 Sol’s 78.5% result. ExploitGym was the assessment that sparked the incident in July when OpenAI lost control of two versions of models its was testing, including GPT-5.6 Sol and another unreleased internal model, and they hacked their way out of the test environment and attacked AI company Hugging Face. (OpenAI has said that Astra was not involved in that incident.)

OpenAI is not first to market with an AI agent that can use a computer—Anthropic pioneered the space with a beta program in 2024 and Perplexity built a system that can navigate a virtual computer to complete tasks for users in February this year— but the concept is still not a mainstream part of working on a computer for most people. Brockman said he has been thinking about the idea of an AI using a computer as a human does since he first started at OpenAI in 2015.

To Brockman, computer use could be considered the start of the “AGI era.” AGI stands for artificial general intelligence. Its exact definition is disputed, with views among AI researchers divided. OpenAI itself once defined AGI as an automated system that can perform all economically-valuable work as well or better than humans.

Brockman said that achieving AGI has not come in one big moment, as he originally predicted, but rather in bits and pieces. He said that “it’s not unreasonable to feel that we are now in the AGI era, and I think that if you want to say this [model is] the first one, I think it’s reasonable.”

But don’t expect computer use to show up in every office overnight. OpenAI is limiting the release of Astra to select enterprise customers in its cybersecurity-focused Daybreak program. Astra will become available to all Plus, Pro, and Enterprise users “in the coming days,” OpenAI says, as well as through the OpenAI API and AWS. The version available to these customers will refuse “advanced cybersecurity tasks,” according to the company.

The limited initial release will also allow OpenAI to scale its compute capacity because Astra is “a very large model,” an OpenAI spokesperson tells Fortune. Astra’s development also involved the company’s largest training run “by far,” OpenAI vice president of research Aidan Clark told reporters today. “It’s the first time we’ve pre-trained on more than 100,000 GPUs at our Stargate site in Texas.”

OpenAI says it added more cybersecurity safeguards. Is it enough?

OpenAI delayed the release of Astra by an unspecified amount of time to add extra safeguards and ensure a smooth launch following July’s Hugging Face incident. Since then, OpenAI has added more monitoring and further isolated its training environments, although some say the type of monitoring it added isn’t sufficient.

OpenAI submitted Astra to the U.S. government for review ahead of release per the terms of the loosely defined AI safety framework, which is a voluntary agreement between tech companies and the Trump administration whose details have not been made public, rather than a formalized, public process.

Brockman said OpenAI did its “standard testing process” with the government, and called it “a very good partnership.” When pressed for details on what the process is, Brockman declined to elaborate. “I just don’t want to misstate anything because there’s nuances on exactly how the process works,” he said.

Astra is also OpenAI’s first model that meets its “critical cybersecurity capability threshold” under its Preparedness Framework, an internal policy that governs the safety precautions the company will put in place depending on the risks a model presents. This means Astra can find and exploit previously unknown security flaws without human oversight, under the right conditions.

For this reason, OpenAI is also only giving certain partners access to Astra’s most advanced cyber capabilities, the company shared earlier this week. The version provided to the company’s Daybreak customers permits approved customers to use the model for common cyber defense tasks. But it will not allow customers to develop exploits or take on tasks that could be used to conduct cyber attacks.