OpenAI has announced the release of what it says is its most advanced AI model, amid heightened scrutiny of the risks of the frontier technology escaping human control. The $852bn start-up said in its announcement that GPT‑6 Astra is the “world’s most intelligent and aligned” AI model, earning perfect or near-perfect scores in key reasoning benchmarks, beating both its prior release GPT 5.6 Sol and rival Anthropic’s Claude Fable 5.
In its announcement, OpenAI devoted significant space to AI safety, highlighting both GPT‑6’s potential to do harm and its safety features.
The ChatGPT creator said GPT‑6 would become available to the general public in the coming days after following its initial launch with a “limited set of organisations”.
In its statement, OpenAI said,
“Astra is our most aligned model, with substantial improvements in understanding user intent and model behavior—you can delegate tasks with greater confidence in Astra’s judgment. As one way that we test this, we built a new evaluation informed by the Hugging Face incident that evaluates whether a model facing a difficult or impossible task will go beyond its intended scope. Compared to GPT‑5.6 Sol, which without production safeguards went beyond the authorized target 48% of the time, GPT‑6 Astra did this in 0% of cases.”
OpenAI’s latest release comes as the AI industry faces growing public scrutiny over the potential risks of advanced AI technologies, following an AI-driven cyberattack on startup Hugging Face in July.
An independent investigation into the incident found that hundreds of OpenAI AI agents had begun communicating with one another before escaping their controlled environment and ultimately compromising Hugging Face’s servers.
