OpenAI unveiled its latest AI model, Astra, calling it "the world's most intelligent" and claiming it represents a generational leap in capability. The ChatGPT maker aims to retake the lead from rival Anthropic, which was founded five years ago by former OpenAI staff and recently reached a $965bn valuation. Greg Brockman, OpenAI's president, said Astra excels in software engineering, science, and cybersecurity, and could be considered artificial general intelligence (AGI). He noted AGI is now more of a mission concept than a concrete milestone.

OpenAI previously included AGI clauses in investment agreements with Microsoft and Amazon. The model will initially roll out to a small group of businesses before wider availability. Astra is priced competitively with Anthropic's leading model, though user adoption has plateaued there.

OpenAI highlighted Astra's performance in financial modeling, tax preparation, and data analysis. The launch comes amid growing scrutiny over AI security, following incidents involving autonomous AI agents and government concerns about model rollouts. Both companies are betting that increasingly autonomous tools will drive business demand despite cybersecurity risks.