GigaWorld

GigaWorld is a project series by OpenGigaAI that develops World Action Models (WAMs) for robot policy learning. WAMs jointly model actions and future visual observations, using future scene evolution as dense supervision for physically grounded action generation. The series includes GigaWorld-0 (data engine), GigaWorld-Policy (action-centered WAM), GigaWorld-Policy-0.5 (efficient MoT-based WAM), and GigaWorld-1 (policy evaluation roadmap).