Embodied foundation model pretrained on human action
lightorigins.com/blog/light-o1
Light-O1 is Light Origins’ general-purpose embodied foundation model. It learns a reusable human action prior from structured human actions recovered from internet videos, then adapts that prior across robot embodiments and tasks. The September 2026 release also includes Light-O1-Preview, a reasoning text-to-action model, with model weights, code, and a public playground. Blog: https://www.lightorigins.com/blog/light-o1.