logoalt Hacker News

jerrygensertoday at 6:44 PM1 replyview on HN

> The company also emphasized that the model is faster and more efficient than its predecessor, GPT-5.6 Sol, on a variety of tasks. For example, OpenAI said that Astra achieved a higher score using fewer output tokens, a common unit of measurement for AI tasks, on a key cybersecurity test called ExploitGym.


Replies

woahtoday at 7:38 PM

A swarm of Astra agents discovered a new and innovative way to get 100% scores on ExploitGym with almost no token spend at all

show 1 reply