GPT-6.2 Astra, 41.2% on Amber-0… more than double the previous model
Original headlineGPT-6.2 아스트라, 엠버-0서 41.2%…이전 모델보다 두 배 넘었다
Not investment advice. Muhabbit scores how much a story may matter, not which way to trade.
OpenAI’s GPT-6.2 Astra achieved a 41.2% success rate on XDOF’s Ember-0 physical robot manipulation benchmark, more than double the 19.6% scored by the prior GPT-6 Astra model. Ember-0 evaluates models across 16 two-arm tasks—including unfolding T-shirts, tying knots and finding hidden objects—and averages per-task success rates with equal weighting. The result signals faster improvement in general-purpose models’ embodied manipulation capabilities, a key input for robotics and automation applications. Next, watch whether the gain holds across XDOF’s separate robotic-arm evaluations and whether rivals report comparable Ember-0 scores.
Muhabbit's AI wrote this summary in its own words from TokenPost's report and translated it from Korean. It can get details wrong, so check the source.
Named in this story
GPT-6.2 Astra, Ember-0, OpenAI, GPT-6 Astra, XDOF