OpenAI has unveiled GPT-6 Astra, calling it its most intelligent and aligned model yet. The first wave is reaching a limited set of organizations, with access expanding over the following days to ChatGPT Plus, Pro, Business, and Enterprise, the OpenAI API, Microsoft Azure, and Amazon Web Services Bedrock. Astra Pro is also coming to Pro, Business, and Enterprise plans, while enterprise administrators must enable access for their workspaces.
Astra targets computer use and professional work across browsers, email, calendars, CRM systems, documents, spreadsheets, presentations, software, scientific analysis, websites, and games. OpenAI reports 72.6% on OSWorld 2.0 at about 40 minutes per task, compared with 65.7% and about 75 minutes for GPT-5.6 Sol. An updated Codex harness helped Astra complete Mind2Web work 1.9 times faster than the current Sol experience.
OpenAI is positioning Astra as its strongest software engineering model, reporting 57.9% on Terminal-Bench 4.0 and 74.1% on DeepSWE v1.1. A new experimental Codex context feature can retain notes across context windows and search earlier requirements and test results. Astra can continue working while asking asynchronous questions, make sensible assumptions, and when consequential choices require user input.
Cybersecurity is the sharpest edge of the release. Astra reached OpenAI’s Critical capability threshold and scored 100% on ExploitBench without production safeguards. It found and used two previously unknown zero-day vulnerabilities during evaluation, which OpenAI says it is disclosing to maintainers. The launch model will refuse advanced requests such as creating proof-of-concept exploits, while OpenAI Daybreak is set to widen access for defensive work. In one impossible-task evaluation, Astra never went beyond its authorized target, though its written reasoning proved harder to monitor than GPT-5.6 Sol.
For OpenAI, Astra connects ChatGPT, Codex, its API business, and distribution through two major cloud platforms in one release. The API model is gpt-6-astra, priced at $10 per million input tokens and $50 per million output tokens. Fast mode offers up to twice the speed at twice the price. The company frames Astra as a broad step across knowledge work, coding, science, computer use, and security.