Figure releases Helix 2.5, enabling zero-shot household task completion

Humanoid robotics firm Figure has unveiled its new neural network, Helix 2.5. The robot successfully performed long-horizon tasks—such as tidying living rooms, folding towels, and making beds—in 30 previously unseen homes without local data collection or fine-tuning. Pre-trained on the Index human behavior dataset, Helix 2.5 achieved a 56% zero-shot success rate in blind tests, compared to 9% for models trained from scratch. Figure noted that Index adds 35 minutes of human behavior data per second and the company has committed $3.5 billion in compute resources for training.