OpenAI has announced GPT-6 Astra, claiming it is "the most intelligent and aligned model in the world." The new frontier model follows GPT-5.6 Sol, Terra, and Luna, released just under two months ago. GPT-6 Astra excels at computer-use, browsing, software engineering, cybersecurity, science, and general professional tasks. A demo video shows Astra handling 3D modeling, building slideshows, and managing multi-step workflows across domains. The model scored 98.6% on ARC-AGI-3, 57.7% on Terminal Bench 4.0, and 59.3% on the Agent's Last Exam, outperforming GPT-5.6 Sol. In cybersecurity benchmarks, Astra achieved a perfect score on Exploit Bench and solved 99.2% of SRE-Bench tasks within four attempts. OpenAI highlights improved alignment and safety features, including protections against advanced cybersecurity tasks and better resistance to jailbreaks. The model is rolling out to a limited set of organizations today, with broader availability planned for ChatGPT Plus, Pro, Business, and Enterprise users, as well as through the OpenAI API and AWS. Pricing is set at $10 per million input tokens and $50 per million output tokens.

OpenAI President Greg Brockman emphasized the shift toward evaluating models by price per task rather than token costs. The launch raises questions about responsible deployment, given Astra's advanced capabilities in sensitive areas like cybersecurity. OpenAI asserts built-in safeguards will help monitor misuse and ensure alignment with intended use cases.