Unitree's UnifoLM-WLA-1.0: one 6B model for 64 whole-body humanoid tasks
WebAssemblyMan · reddit · 2026-10-03
Unitree released UnifoLM-WLA-1.0, a general-purpose humanoid foundation model:
- Scale: 6B parameters, trained on 2,500 hours of real robot data; one model handles 64 tasks (10 whole-body + 54 tabletop), supports parallel grippers and two dexterous hands, and beats many open-source models on embodied benchmarks.
- Architecture: built on UnifoLM-ER-1 (a Qwen3-VL-based embodied reasoner), adds future dynamic region prediction via optical flow + VQ-VAE, discretizes actions (end-effector/hand/lower body) with residual VQ, then an MMDiT action expert for continuous control.
- Demos: on Unitree G1 making the bed, loading the washing machine, folding clothes, sorting objects.
The poster calls it one of the most complete open attempts at a whole-body VLA so far.
More from Embodied
- After 1000 runs of home robot Rooma, author loses faith in RL — OnlineInference · 2026-10-03
- MolmoMotion: AI2's 4B VLM forecasts 3D point trajectories from language instructions — rsasaki0109 · 2026-10-03
- Video shows Waymo robotaxi navigating city streets and merging onto freeway — reed · 2026-10-03
- OpenAI's Dot hardware is legit despite a broken demo, early review says — altryne · 2026-10-03
- ALOHA creator quits Stanford PhD for Sunday Robotics; LeRobot's Cadene builds UMA — FinanceYF5 · 2026-10-03
- Dev turns ESP32-S3 into a Muse gadget early using the official Muse SDK — alexandr_wang · 2026-10-03