LeRobot v0.6.0 released with world models and new VLA models

Hugging Face announced the release of LeRobot v0.6.0 on July 7, 2026. The version adds three world model policies — VLA-JEPA, FastWAM and LingBot-VA — and new VLA models, including GR00T N1.7, MolmoAct2, EO-1, EVO1 and Multitask DiT. It also introduces a reward model API that includes Robometer and TOPReward, six simulation benchmarks, the lerobot-rollout CLI, FSDP training and cloud training capabilities.

Robot learning researchers and developers can test policies that use future predictions, reward models that assess progress and success, and evaluations in simulated environments within the same framework. They can also more easily compare methods and check how to standardize data recording and training conditions before deployment on real robots.