Figure's Helix 2.5 Lets Humanoid Robots Perform Housework in Unseen Homes
Humanoid robot company Figure released Helix 2.5 on September 18, 2026, a model designed to let robots carry out household tasks in new environments without prior adaptation. The company rented 30 homes in the San Francisco Bay Area, where robots powered by the model began tidying the spaces without additional training.
According to Figure, Helix 2.5 was built to answer whether a humanoid can enter a home it has never seen and immediately work autonomously. The model was pretrained on Index, a global dataset of human behavior. From that single foundation model, the company tuned three behaviors: tidying a living room, folding towels, and making a bed. The tasks cover movement, rigid and deformable object manipulation, bimanual coordination, and active perception.
In tests, Index pretraining significantly improved zero-shot whole-body generalization for humanoid robots. With task-specific data, architecture, training, and evaluation unchanged, success rates rose from 8% to 56%. Compared with the previous neural network Helix 02, Helix 2.5 used half the task-specific data to generalize the behavior to 30 previously unseen environments.
Figure said Helix 2.5 is the first proof that whole-body intelligence can be learned from human experience and transferred to new scenes, without rebuilding it each time.
