Models

pi0.5 OpenPI pi05_libero VLA policy running closed-loop LIBERO/MuJoCo rollouts, with DROID action API checks as a secondary profile. vlacloudopenpilibero

11.2Kpulls 2tags 3 months agoupdated

Gemma 4 Multimodal VLM family for robot scene understanding, tool calls, coding, and local high-memory Apple Silicon experiments. vlmvisionlocalcloud31b

5.5Mpulls 29tags 2 weeks agoupdated

Cosmos 14B World-model runtime for video-to-world, text-to-world, and policy evaluation workloads; measured at 7.724s for a 17-frame 480p one-step H200 smoke. worldcloudgpu14b

7.6Kpulls 8tags 4 months agoupdated

Ctrl-World Controllable generative world model for robot manipulation rollouts, unseen DROID-style tasks, and pi0.5 policy-in-the-loop evaluation. worldtrainingdroidcloud

100pulls 1tag 5 months agoupdated

BAGEL World Model Public BAGEL-7B-MoT image model benchmarked on one Nebius H200, with fastest valid reduced subgoal inference at 2.138s. worldcloudh200subgoal image

8.6Kpulls 9tags 3 months agoupdated

php-parkour / G1 Depth-based Unitree G1 parkour policy running in browser MuJoCo WASM, with the mobile route using custom JS policy and depth runners. simbrowserg1js runtime

2pulls 1tag 4 weeks agoupdated

SimToolReal / iiwa_sharpa Object-centric dexterous manipulation policy for the LBR iiwa 14 and Sharpa HA4, packaged as a local browser sim. simbrowseriiwasharpa

213pulls 1tag yesterdayupdated