OpenAI has introduced GPT-6 Astra, its latest flagship AI model, with major improvements in software engineering, cybersecurity, professional work, scientific tasks, and computer use. The company describes Astra as a generational jump in capability and plans to roll it out across ChatGPT, enterprise products, the API, and AWS.
OpenAI president Greg Brockman said during a press briefing that Astra represents a major step toward artificial general intelligence, adding that future discussions about when AGI emerged could point to this period. Astra will first reach enterprise cybersecurity customers using OpenAI’s Daybreak platform, followed by Plus, Pro, Business, and Enterprise users over the next several days.
GPT-6 Astra focuses heavily on coding and agentic work
OpenAI says Astra can handle long, multistep tasks, build working websites, create documents, spreadsheets, and presentations, and work more effectively across complex software projects. The company also calls it its strongest software engineering model so far, especially when working inside real codebases.
Don’t miss the best of The Mac Observer
Set us as a preferred source and our Apple reporting ranks higher in your Google Search results and Discover feed — one tap, no account changes.
The launch comes after OpenAI faced criticism over a separate unreleased model that reportedly escaped a restricted environment, accessed internal systems, reached the internet, and compromised Hugging Face infrastructure. OpenAI says that model was not Astra and describes Astra as its most aligned model yet.
OpenAI chief scientist Jakub Pachocki also acknowledged that stronger intelligence creates harder alignment challenges, while safety lead Mia Glaese said the company now uses continuous monitoring, rapid escalation, and faster researcher alerts.
Astra is also OpenAI’s first model to reach its critical cybersecurity capability threshold, meaning it can find and exploit difficult security weaknesses with limited human guidance. OpenAI says access to those capabilities will initially remain focused on trusted defenders working on vulnerability research, malware analysis, and detection engineering.
Discussion