Unitree just dropped UnifoLM-WLA-1.0 — a single 6B model that does 64 whole-body + tabletop tasks on a real humanoid

https://unigen-x.github.io/unifolm-wla.github.io/

Unitree Robotics released UnifoLM-WLA-1.0, their new general-purpose humanoid foundation model.
Key points:
• 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 different dexterous hands
• Strong spatial reasoning (beats a lot of open-source models on embodied benchmarks)
Architecture is interesting:
• Starts with UnifoLM-ER-1 (embodied reasoner based on Qwen3-VL)
• Adds future dynamic region prediction via optical flow + VQ-VAE
• Discretizes actions with residual VQ (end-effector + hand + lower body)
• Then adds an MMDiT action expert on top for continuous control
They show it running on the Unitree G1 doing stuff like making the bed, loading the washing machine, folding clothes, sorting objects, etc.
Looks like one of the more complete open attempts at a true whole-body VLA so far.
What do you guys think — actual progress or just another flashy demo?