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 AWE 3.5, a generalist embodied AI model(3 posts)→
More from Embodied
- After 50+ robotaxi rides, user ranks Tesla Cybercab the best rider experience — EricETesla · 2026-10-16
- Robot Startup Ditches VLAs for Frontier LLMs in a Customer Deployment — ihorbeaver · 2026-10-16
- CMU's Andrea Bajcsy on what safety means for generalist robots, beyond collision avoidance — CSProfKGD · 2026-10-16
- Codex drives a physical CAN-bus motor, eyeing a fully Codex-built robot — OpenAIDevs · 2026-10-16
- Berkeley paper uses the Midas myth to rethink robotics beyond human imitation — begusgasper · 2026-10-16
- Team uses Claude Opus 5.5 to reproduce core claims of 10 recent robotics papers end to end — teortaxesTex · 2026-10-16