OpenAI launches GPT-6 Astra, its most intelligent and aligned model yet

The new frontier model tops ARC-AGI-3 and ExploitBench, rolls out to ChatGPT, the API, and major clouds, and is now available in GitHub Copilot and Microsoft Foundry.

OpenAI has released GPT-6 Astra, a frontier model it is billing as "the world's most intelligent and aligned model." Astra saturates several of the hardest benchmarks — 98% on FrontierMath Tier 4, 99.9% on ARC-AGI-3, and a perfect 100% on ExploitBench — and OpenAI says it completes real computer-use tasks roughly 47% faster than its predecessor, GPT-5.6 Sol.

The rollout is staggered. Astra is going out now to a limited set of organizations, then over the coming days to all ChatGPT Plus, Pro, Business, and Enterprise users, plus the OpenAI API, Microsoft Azure, and AWS Bedrock. Standard API pricing is set at $10 per million input tokens and $50 per million output tokens.

The model is already spreading across the developer ecosystem. GitHub made GPT-6 Astra generally available in Copilot Pro+, Max, Business, and Enterprise plans, describing it as built for "long-horizon, autonomous coding and agentic tasks." Microsoft likewise announced the model is generally available in Microsoft Foundry for enterprise customers, framing the next era of enterprise AI around how well a model can "work for and with you" rather than chat alone.

Safety is front and center: OpenAI says Astra refuses advanced cybersecurity tasks such as generating proof-of-concept exploits, and it plans to widen defensive access through its Daybreak program in the coming weeks.

Sources: GitHub | Microsoft