Where physical AI gets built.

A compact 450M-parameter VLA from Hugging Face, pretrained on community-collected LeRobot datasets and sized for fine-tuning and inference on consumer GPUs.

LeRobot's port of the Physical Intelligence pi0.5 base model, which is co-trained on web and multi-robot data to generalize better to unfamiliar homes and objects.

A 3B cross-embodiment robot foundation model from NVIDIA that combines a vision-language backbone with a diffusion-transformer action head and can be post-trained on a new robot with a modest amount of data.

A 7B vision-language-action model on a Llama 2 backbone that maps one camera image and a text instruction to discretized 7-DoF end-effector actions, trained on Open X-Embodiment episodes.

A 93M-parameter transformer generalist policy with a diffusion action head, conditioned on language or goal images and intended to be fine-tuned to new robots, sensors, and action spaces.

LeRobot's port of the Physical Intelligence pi0 base checkpoint, a PaliGemma-based VLA that outputs continuous action chunks through a flow-matching action expert.

The small size of the Moonshine Streaming family, an English-only encoder-decoder recognizer that uses a sliding-window encoder so it can transcribe live audio with low latency on edge hardware.

GGML-format conversions of the OpenAI Whisper checkpoints, from tiny to large, for the whisper.cpp C/C++ inference engine.

An 82M-parameter text-to-speech model built on StyleTTS 2 components, shipped with a set of preset voices across several languages.

The official collection of ONNX voice models for the Piper offline TTS engine, covering many languages at several size and quality levels.

A 100M-parameter CPU-oriented TTS model from Kyutai with voice cloning from a reference clip and support for English plus five other European languages, designed to run faster than real time using two CPU cores.

ONNX exports of the Silero voice activity detector in fp32, fp16 and 8-bit forms (the int8 file is about 0.6 MB) that mark which short audio frames contain speech.

Meta's self-supervised ViT-L video encoder, pretrained by predicting masked video regions in latent space rather than in pixels.

Compact variant of Meta's promptable segmentation model that returns object masks from points or boxes and can track them across video frames.

Open-vocabulary detector that draws boxes around objects named in a free-text prompt, without a fixed class list.

CLIP-based zero-shot detector from Google that localizes objects from text queries or from an example image crop.

Small real-time DETR-style detector for the 80 COCO classes that refines box edges as distributions for tighter localization.

Lightweight monocular model that predicts a relative depth map from a single RGB frame.

Fifty real pick-and-place episodes from a low-cost SO-100/SO-101-class arm, recorded at 30 fps with a side camera and a top-down camera, and used as the SmolVLA fine-tuning example.

A filtered collection of 128 SO-100 datasets from 55 community contributors (about 11k episodes and 5.1M frames), with hand-cleaned task descriptions, released as the pretraining subset for SmolVLA.

About 200 human demonstrations in a 2D simulation where a round pusher slides a T-shaped block onto a target outline, stored in LeRobot v3 format.

Fifty human-teleoperated simulated episodes of a two-arm ALOHA robot handing a cube from one gripper to the other, recorded at 50 fps.

LeRobot v3 conversion of DROID: about 95k real Franka teleoperation episodes recorded in many different scenes, with around 50k distinct language instructions.

LeRobot-format copy of the LIBERO simulated Franka Panda benchmark, with about 1.7k demonstrations across its 40 language-conditioned tasks.