{"version":1,"type":"story","url":"https://digestai.news/story/figure-unveils-helix-2-5-humanoid-robot-achieving-56-zero-shot-success","json":"https://digestai.news/story/figure-unveils-helix-2-5-humanoid-robot-achieving-56-zero-shot-success.json","markdown":"https://digestai.news/story/figure-unveils-helix-2-5-humanoid-robot-achieving-56-zero-shot-success.md","slug":"figure-unveils-helix-2-5-humanoid-robot-achieving-56-zero-shot-success","headline":"Figure unveils Helix 2.5 humanoid robot, achieving 56% zero-shot success in 30 homes","summary":"Figure announced its new humanoid neural network, Helix 2.5, on September 17, 2026. The robot was pretrained on the company’s Index dataset of human behavior and evaluated in 30 Bay Area homes that were completely unseen during training. In a controlled comparison, the Index‑pretrained policy succeeded on 56% of zero‑shot trials, while an otherwise identical policy trained from scratch managed only 9% success.\n\nHelix 2.5 performed three whole‑body tasks—tidying living rooms, folding towels, and making beds—using only the furniture and objects present in each home. The system matched the performance of a prior Helix 02 policy that required data collected on‑site, while using half the adaptation data. Figure also reported a scaling law: increasing Index data eight‑fold reduced held‑out prediction loss predictably, and the company has committed $3.5 B of compute to training the robot. Index now generates roughly 35 minutes of new human video per second, having amassed 264 000 app downloads, 16 million videos, and $15 million paid to contributors.","keyPoints":["Helix 2.5 achieved 56% zero‑shot task success across 30 unseen Bay Area homes, versus 9% for a scratch‑trained policy.","The robot was pretrained on Figure’s Index dataset, which now collects ~35 minutes of human video per second and has $3.5 B compute allocated.","Helix 2.5 matched the performance of a home‑specific policy while using half the adaptation data, demonstrating a scaling law for data efficiency."],"whyItMatters":"Shows that large‑scale human behavior pretraining can give humanoid robots generalizable skills without site‑specific data, lowering deployment costs and accelerating physical AI commercialization.","category":{"slug":"robotics","name":"Robotics & Physical AI","url":"https://digestai.news/category/robotics"},"entities":{"companies":["Figure"],"models":["Helix 2.5","Helix 02"],"people":[]},"firstPublishedAt":"2026-09-17T18:46:02Z","updatedAt":"2026-09-17T18:46:02Z","sourceCount":1,"hasPrimarySource":false,"sources":[{"outlet":"Unite.AI","title":"Figure Introduces Helix 2.5, Tested Zero-Shot in 30 Unseen Homes","url":"https://unite.ai/figure-introduces-helix-2-5-tested-zero-shot-in-30-unseen-homes","publishedAt":"2026-09-17T18:46:02Z","type":"press","primary":false,"lead":true}],"sourceNotes":null,"discussions":[],"thread":null,"cite":{"text":"Digest AI, \"Figure unveils Helix 2.5 humanoid robot, achieving 56% zero-shot success in 30 homes\", 17 September 2026, https://digestai.news/story/figure-unveils-helix-2-5-humanoid-robot-achieving-56-zero-shot-success","publisher":"Digest AI","title":"Figure unveils Helix 2.5 humanoid robot, achieving 56% zero-shot success in 30 homes","datePublished":"2026-09-17T18:46:02Z","url":"https://digestai.news/story/figure-unveils-helix-2-5-humanoid-robot-achieving-56-zero-shot-success"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}