OpenAI aims to achieve AGI by year-end, with Astra tackling advanced math and desktop tasks

OpenAI believes it will build artificial general intelligence before the ball drops on 2027. CEO Sam Altman laid out the timeline in a TIME cover story, projecting that the company will have an internal system qualifying as AGI by December 31, 2026. Chief Research Officer Mark Chen put a number on it: OpenAI is 80% of the way there. For a company valued at roughly $1 trillion with over 1 billion active ChatGPT users, that remaining 20% might be the most consequential stretch of R&D in the history of computing. What Astra actually does The vehicle for this ambition is Astra, OpenAI’s forthcoming model family designed to coordinate up to 16 AI agents simultaneously. Those tasks aren’t trivial. Astra is built to tackle advanced mathematical research and navigate desktop software at what OpenAI describes as “superhuman speed.” The system has already notched a genuinely impressive feat: solving 10 previously unsolved theoretical computer science problems. OpenAI also claims it has already surpassed its internal benchmarks for automating entry-level AI research work. A June 2026 roadmap had originally targeted significant internal use of automated AI researchers by March 2028. The de...