Amazon Open-Sources DreamFAR, a Real-Time World Model That Nearly Doubles Robot Performance

Amazon FAR, with Cornell, Maryland and UPenn, released DreamFAR, an open-source real-time interactive world model for robot manipulation that runs at 10+ FPS with an online demo. Trained purely on native robot data, it enables persistent object-state consistency in long-horizon interactions and nearly doubles π₀'s real-robot performance.

2026-10-30 ~ 2026-10-30 · 3 related posts

• Amazon FAR's DreamFAR: a world model trained from scratch on robot data only, 1,000+ tasks — _krishna_murthy · 2026-10-30
• Amazon's DreamFAR world model nearly doubles π₀ real-world performance for robot training — CyberRobooo · 2026-10-30
• Amazon's DreamFAR: Real-Time Robot World Model at 10+ FPS, Playable Online — CyberRobooo · 2026-10-30