Skip to content
Every·· 1 d agoAI score58

Vibe Check: GPT-6.2 Astra and Sol Clear Tasks That Stumped Models Earlier This Year

Vibe Check: GPT-6.2 Astra and Sol Leave Us Wanting More

AI summary

Every's Vibe Check evaluates OpenAI's GPT-6.2 Astra and GPT-6.2 Sol on tasks that no model could handle well at the start of this year, including a themed training curriculum, a minute-by-minute show plan, and executive exercises with realistic data. Head of evals Mike Taylor says both new models handle all of them, and that their predecessors already performed well, with Codex having largely saturated these tasks. He calls the progress great to see but says it leaves him wanting more.

Source:Every · every.to