A robot made beds in 30 homes it had never entered.
Figure's new Helix 2.5 model finished 56% of household chores in homes it had never seen. Without the new pretraining the score was 9%, measured over 420 attempts at beds, towels and tidying. Figure rented the Bay Area houses and collected no training data inside them first.
Why it mattersHome robots have failed for decades because every house is laid out differently. A model that walks in cold is the first sign that is changing.
▶ Related video: Helix 2.5 30-Home Generalization
Read in the app — free, in 9 languages
Related stories
Mathematicians used AI to close the last gap in a 40-year hunt.2026-09-23New Windows malware lets four chatbots vote on its next move.
2026-09-23OpenAI's million-dollar proof leaned on a force nature never makes.
2026-09-23An AI broke an Enigma message unsolved since 2005.
2026-09-22Alphabet just gave away the software layer under factory robots.
2026-09-22