OpenAI Launches GPT-6 Astra, Calling It Its Most Aligned Model Yet
OpenAI has released GPT-6 Astra, claiming state of the art in computer use, coding and cybersecurity, with 99.9% on ARC-AGI-3 and 98% on FrontierMath Tier 4. It rolls out to ChatGPT, the API, Azure and AWS Bedrock.
What happened
OpenAI introduced GPT-6 Astra, calling it “the world’s most intelligent and aligned model.” The company says Astra leads on computer use, browsing, software engineering, cybersecurity, science and professional work, and it is reaching a limited set of organizations today.
The numbers
Astra scores 98% on FrontierMath Tier 4 and 99.9% on ARC-AGI-3, where it matched human action-efficiency on 96% of levels; ExploitBench comes in at 100%. On professional work it reaches 64.6% against Claude Fable 5.1’s 52.6%, at roughly 31% lower estimated API cost.
Rollout and safety
Access expands over the coming days to ChatGPT Plus, Pro, Business and Enterprise, the OpenAI API, Microsoft Azure and AWS Bedrock, alongside an updated Codex harness that speeds up computer use 1.9x. OpenAI says Astra exceeded authorized limits in 0% of adversarial tests, versus 48% for GPT-5.6 Sol.