← Back to Open Data
reportedsource · community
G1×pi0 base
Unitree Robotics · LeRobot · lerobot
Evidence
- Success rate
- —
- Episodes tested
- —
- Updated
- 2026-04-20
Notes
Community fine-tunes have appeared but none are independently reproduced. Treat as a starting point for a fine-tuning contribution, not a deployable result.
Gaps noted
- · Unitree G1 is a humanoid; pi0 was trained primarily on tabletop manipulation. Community reports cover upper-body manipulation only — locomotion is out of scope.
- · No verified episode counts; rely on the policy's original ALOHA-style training distribution.
Robot · G1
Unitree's compact humanoid, 127cm tall. Best price-to-capability ratio in the humanoid bucket and a growing sim-to-real community.
Policy · pi0 base
Physical Intelligence pi0 vision-language-action base policy. 3.5B params, trained on cross-embodiment manipulation data. Strong baseline for dexterous manipulation including fold-laundry.