Research
Figure announces Helix 2.5, a humanoid control model pretrained on its Index dataset
On September 17, 2026, Figure announced Helix 2.5, a neural network for whole-body control of a humanoid robot on long tasks in homes it has not visited. Figure says Helix 2.5 was pretrained from random initialization entirely on Index, its dataset of human behavior, and that a single base model was then adapted to tidying living rooms, folding towels, and making beds. Figure tested these behaviors in 30 Bay Area homes in which it says no data had been collected. The announcement does not say whether Helix 2.5's weights or code will be released.1
Related catalog entries
Sources
Share or cite this page
- X (opens X)
- Bluesky (opens Bluesky)
- LinkedIn (opens LinkedIn)
- Reddit (opens Reddit)
- Hacker News (opens Hacker News)
- Email (opens your email app)
- RSS updates
Cite this page
APA style
USASI. (2026, October 13). Figure announces Helix 2.5, a humanoid control model pretrained on its Index dataset. United States of America Superintelligence. https://unitedstatesofamericasuperintelligence.com/news/figure-announces-helix-2-5/
BibTeX
@misc{usasi_news_figure_announces_helix_2_5,
author = {{USASI}},
title = {Figure announces Helix 2.5, a humanoid control model pretrained on its Index dataset},
year = {2026},
month = oct,
howpublished = {\url{https://unitedstatesofamericasuperintelligence.com/news/figure-announces-helix-2-5/}},
note = {United States of America Superintelligence. Last updated 2026-10-13}
}The date is when this page was last updated. Add the date you read it if your style needs one.