TARS AWE 3.5 Shows Cross-Scene Generalization with Million-Hour Training
heyshrutimishra · x · 2026-08-19
TARS unveiled its embodied AI foundation model, AWE 3.5, demonstrating generalization across diverse tasks like smartphone packing, backpack packaging, screw sorting, and cable plugging.
Technical Highlights:
- Uses a "Born as One" architecture, natively integrating vision, touch, geometry, and motion.
- Trained on over 1 million hours of human-centric data.
- The first native model to complete both pre-training and post-training.
Background: Their A1 robot previously set a Guinness World Record for wire harness assembly. Co-founder Dr. Wenchao Ding was named to MIT Technology Review's TR35 China 2025 list.
Related event: TARS Unveils Embodied AI Model AWE 3.5 with Cross-Task Generalization(2 posts)→
More from Embodied
- ROS robot with 3D stereo camera follows you and plays MTV videos — chrismatthieu · 2026-08-20
- World Humanoid Robot Race: Pupils in 2025, Athletes by 2026 — iamfakhrealam · 2026-08-19
- Opinion: Humanoid robots as future armies — devanshmehta · 2026-08-19
- YC S26 startup Grip uses foundation models to sort unsortable waste — ycombinator · 2026-08-19
- Synchron's Stentrode BCI works via blood vessels, no brain surgery; Apple now accepts brain signals as native input — import_jmr · 2026-08-19
- QNX partners with Momenta for safety-certified autonomous driving platform — pdamodaran · 2026-08-19