OpenAI’s newest model family, known as Astra, is doing things that would have sounded like science fiction two years ago. It cracked ten longstanding open problems in mathematics and theoretical computer science, producing verified proofs at a compute cost of roughly $2,000. For context, some of those problems had stumped human researchers for decades.
First publicly named on August 1, 2026, Astra represents a substantial leap beyond the GPT-5.6 family that OpenAI released on July 9, 2026. Whether the company officially brands it as GPT-6 remains an open question, but the capabilities gap between Astra and its predecessor is wide enough that the label feels earned.
What Astra can actually do
The math results are impressive on their own, but the cybersecurity milestone might matter more in practical terms. On September 1, 2026, Astra became the first model to meet the “Critical” cybersecurity capability threshold under OpenAI’s Preparedness Framework. That means it can autonomously discover and exploit previously unknown vulnerabilities.
OpenAI is keeping that particular feature locked down. The advanced cyber capabilities are restricted to a small group of testers at launch.
