LeRobot v0.5.0: Scaling Every Dimension

Source: https://huggingface.co/blog/lerobot-release-v050 Back to Articles

• TL;DR
• Table of Contents
• Hardware: More Robots Than Ever- Unitree G1 Humanoid - OpenArm & OpenArm Mini - More Robots - CAN Bus Motors
• Policies: A Growing Model Zoo- Pi0-FAST: Autoregressive VLAs - Real-Time Chunking (RTC) - Wall-X - X-VLA - SARM - PEFT Support
• Datasets: Faster Recording, Faster Training- Streaming Video Encoding - 10x Faster Image Training, 3x Faster Encoding - New Dataset Tools
• EnvHub: Environments from the Hub- NVIDIA IsaacLab-Arena
• Codebase: A Modern Foundation
• Community & Ecosystem
• Final Thoughts

With over 200 merged PRs and over 50 new contributors since v0.4.0, LeRobot v0.5.0 is our biggest release yet — expanding in every direction at once. More robots (including our first humanoid), more policies (including the comeback of autoregressive VLAs), faster datasets, simulation environments you can load straight from the Hub, and a modernized codebase running on Python 3.12 and Transformers v5. Whether you’re training policies in simulation or deploying them on real hardware, v0.5.0 has something for you.

https://huggingface.co/blog/lerobot-release-v050#tldrTL;DR

LeRobot v0.5.0 adds full Unitree G1 humanoid support (whole-body control models), new policies –including Pi0-FAST autoregressive VLAs and Real-Time Chunking for responsive inference–, and streaming video encoding that eliminates wait times between recording episodes. The release also introduces EnvHub for loading simulation environments from the Hugging Face Hub, NVIDIA IsaacLab-Arena integration, and a major codebase modernization with Python 3.12+, Transformers v5, and third-party policy plugins.

https://huggingface.co/blog/lerobot-release-v050#table-of-contentsTable of Contents

• LeRobot v0.5.0: Scaling Every Dimension- TL;DR - Table of Contents - Hardware: More Robots Than Ever- Unitree G1 Humanoid - OpenArm & OpenArm Mini - More Robots - CAN Bus Motors - Policies: A Growing Model Zoo- Pi0-FAST: Autoregressive VLAs - Real-Time Chunking (RTC) - Wall-X - X-VLA - SARM - PEFT Support - Datasets: Faster Recording, Faster Training- Streaming Video Encoding - 10x Faster Image Training, 3x Faster Encoding - New Dataset Tools - EnvHub: Environments from the Hub- NVIDIA IsaacLab-Arena - Codebase: A Modern Foundation - Community & Ecosystem - Final Thoughts

https://huggingface.co/blog/lerobot-release-v050#hardware-more-robots-than-everHardware: More Robots Than Ever

LeRobot v0.5.0 dramatically expands the roster of supported hardware — from arms and mobile robots to a full humanoid.

https://huggingface.co/blog/lerobot-release-v050#unitree-g1-humanoidUnitree G1 Humanoid

The biggest hardware addition in this release:full Unitree G1 humanoid support. This is LeRobot’s first humanoid integration, and it’s comprehensive:

• Locomotion: Walk, navigate, and move through environments.
• Manipulation: Perform dexterous object manipulation tasks.
• Teleoperation: Control the G1 remotely with an intuitive teleoperation interface.
• Whole-Body Control (WBC): Coordinate locomotion and manipulation simultaneously for complex, real-world tasks.

The G1 integration represents a major step toward general-purpose robotics within LeRobot — moving beyond tabletop arms into full-body embodied AI. Try it out yourself by following thedocumentation.

https://huggingface.co/blog/lerobot-release-v050#openarm–openarm-miniOpenArm & OpenArm Mini

We’ve added support for theOpenArmrobot and its companionOpenArm Miniteleoperator. OpenArm is a capable robot arm with full LeRobot integration, and the Mini serves as its natural teleoperation device. Both supportbi-manual configurations, enabling dual-arm setups for more complex manipulation tasks. Check it out in thedocumentation.

https://huggingface.co/blog/lerobot-release-v050#more-robotsMore Robots

The hardware ecosystem keeps growing:

• Earth Rover: Our first mobile robot integration, bringing LeRobot to outdoor navigation and ground-level robotics.
• OMX Robot: A new robot arm with configurable gripper settings and calibration support.
• SO-100/SO-101 Consolidation: We’ve unified the SO-100 and SO-101 implementations into a single, cleaner codebase — including bi-manual setups. Less code duplication, easier maintenance, same great robots.

https://huggingface.co/blog/lerobot-release-v050#can-bus-motorsCAN Bus Motors

New motor controller support via CAN (Controller Area Network) bus opens the door to higher-performance actuators:

• RobStride: A CAN-based motor controller for high-torque applications.
• Damiao: Another CAN bus motor controller, expanding the range of compatible hardware.

These additions mean LeRobot can now drive a wider variety of professional-grade actuators beyond the existing Dynamixel and Feetech ecosystem.

https://huggingface.co/blog/lerobot-release-v050#policies-a-growing-model-zooPolicies: A Growing Model Zoo

This release brings six new policies and techniques into LeRobot, pushing the boundaries of what’s possible with open-source robot learning.

https://huggingface.co/blog/lerobot-release-v050#pi0-fast-autoregressive-vlasPi0-FAST: Autoregressive VLAs

Pi0-FASTbrings autoregressive Vision-Language-Action models toLeRobot with FAST (Frequency-space Action Sequence Tokenization). Unlike the flow-matching approach of Pi0, Pi0-FAST uses an autoregressive action expert (based on Gemma 300M) that generates discretized action tokens, enabling:

• FAST tokenization: Actions are tokenized for autoregressive decoding, with a dedicatedFAST action tokenizer.
• Flexible decoding: Configurable temperature and max decoding steps for balancing speed and quality.
• RTC-compatible: Works with Real-Time Chunking (seenext section) for responsive inference.

lerobot-train \
  --policy.type=pi0_fast \
  --dataset.repo_id=lerobot/aloha_sim_insertion_human \
  --policy.device=cuda

https://huggingface.co/blog/lerobot-release-v050#real-time-chunking-rtcReal-Time Chunking (RTC)

Real-Time Chunkingis an inference-time technique fromPhysical Intelligencethat makes flow-matching policies dramatically more responsive. Instead of waiting for a full action chunk to finish before replanning, RTC continuously blends new predictions with in-progress actions, producing smoother and more reactive behavior.

RTC is not a standalone policy — it’s an enhancement that plugs into existing flow-matching policies (Pi0 family, SmolVLA & Diffusion). Configure it via\-\-policy\.rtc\_config\.enabled=true.

This is a game-changer for real-world deployment where latency matters. Read theoriginal paperfor the technical details and ourdocumentation.

https://huggingface.co/blog/lerobot-release-v050#wall-xWall-X

Wall-Xis a new VLA policy built onQwen2.5-VLwith flow-matching action prediction. It combines the strong vision-language understanding of Qwen2.5-VL with a flow-matching head for cross-embodiment robotic control.

pip install lerobot[wall_x]
lerobot-train \
  --policy.type=wall_x \
  --dataset.repo_id=lerobot/aloha_sim_insertion_human

https://huggingface.co/blog/lerobot-release-v050#x-vlaX-VLA

X-VLAbrings aFlorence2-basedVLA to LeRobot. Built on Microsoft’s Florence-2 vision-language model, X-VLA offers an alternative backbone for VLA policies, expanding the diversity of foundation models available for robot learning. Check out thetraining guidefor setup instructions and thebase model.

pip install lerobot[xvla]
lerobot-train \
  --policy.type=xvla \
  --dataset.repo_id=lerobot/bimanual-so100-handover-cube

https://huggingface.co/blog/lerobot-release-v050#sarmSARM

**SARM (Stage-Aware Reward Modeling)**tackles one of the hardest problems in robot learning: long-horizon tasks. Instead of using a single global linear progress signal over the whole episode, it models progress in a stage-aware manner by predicting both the task stage and the progress within that stage. This makes it much easier to train policies for complex, multi-step manipulation tasks. Start experimenting with it by following thedocumentation.

https://huggingface.co/blog/lerobot-release-v050#peft-supportPEFT Support

You can nowfine-tune large VLAs using LoRA(and other PEFT methods) without modifying the core training pipeline. PEFT configuration lives at the policy level, making it straightforward to adapt massive foundation models to your specific robot and task with a fraction of the compute. Learn more reading thedocumentation.

lerobot-train \
  --policy.type=pi0 \
  --policy.peft_config.use_peft=true \
  --dataset.repo_id=lerobot/aloha_sim_insertion_human

https://huggingface.co/blog/lerobot-release-v050#datasets-faster-recording-faster-trainingDatasets: Faster Recording, Faster Training

The dataset pipeline gets major performance improvements in this release, making both data collection and training significantly faster.

https://huggingface.co/blog/lerobot-release-v050#streaming-video-encodingStreaming Video Encoding

Previously, recording a dataset meant waiting after each episode for video encoding to finish.No more.With streaming video encoding, frames are encoded in real-time as they’re captured — meaningzero wait time between episodes. Just finish one episode and immediately start the next.

Streaming encoding also supportshardware encoder auto-detection, so if your system has a GPU-accelerated video encoder, LeRobot will use it automatically:

dataset = LeRobotDataset.create(
    repo_id="my/dataset",
    fps=30,
    video_backend="auto",       # Auto-detect best HW encoder
    streaming_encoding=True,    # Encode in real-time
)

Streaming encoding performance can vary significantly depending on your hardware and recording setup (number of cameras, resolution, etc.). Make sure to review thestreaming video encoding documentationbefore enabling it.

https://huggingface.co/blog/lerobot-release-v050#10x-faster-image-training-3x-faster-encoding10x Faster Image Training, 3x Faster Encoding

Under the hood, we’ve fixed key data access bottlenecks and overhauled image processing:

• 10x faster image training: Improved image transform support and fixed data access bottlenecks that were silently slowing down training.
• 3x faster encoding: Parallel encoding is now the default across all platforms, with dynamic compression levels that adapt to your dataset type (video vs. image), when not using streaming.
• Better CPU utilization: More efficient resource usage during recording and dataset creation.

https://huggingface.co/blog/lerobot-release-v050#new-dataset-toolsNew Dataset Tools

The dataset editing toolkit continues to grow:

• Subtask support: Annotate and query subtasks within episodes for hierarchical task learning.
• Image-to-video conversion: Convert existing image-based datasets to video format for better storage efficiency, with support for multiple episodes per video file.
• More editing operations: Newinfooperation for inspecting datasets, task modification tools, and numerous fixes to existing operations (splitting, merging, feature editing).
• Expose more options: Configurable video codecs, tolerance settings, and metadata buffer sizes for fine-grained control over dataset creation.

https://huggingface.co/blog/lerobot-release-v050#envhub-environments-from-the-hubEnvHub: Environments from the Hub

EnvHubis a new way to use simulation environments in LeRobot: load them directly from the Hugging Face Hub. Instead of installing environment packages locally and wiring up registration, you can now point LeRobot at a Hub repository and it handles everything — downloading the environment code, registering it with Gymnasium, and making it available for training and evaluation.

Hub environments useHubEnvConfig, which downloads and executes remotemake\_envfunctions:

lerobot-train \
  --env.type=hub \
  --env.hub_path="username/my-custom-env" \
  --policy.type=act

This lowers the barrier for sharing custom simulation environments with the community. Package your environment, push it to the Hub, and anyone can train on it. Check out thedocumentationto learn more. Here’s an example to get started:LeIsaac x LeRobot EnvHub tutorial.

https://huggingface.co/blog/lerobot-release-v050#nvidia-isaaclab-arenaNVIDIA IsaacLab-Arena

We’ve integratedNVIDIA IsaacLab-Arena, bringing GPU-accelerated simulation to LeRobot. IsaacLab-Arena provides a collection of manipulation tasks running on NVIDIA’s Isaac Sim, offering massively parallel environment instances for fast reinforcement learning. The integration includes dedicated pre/post-processing steps and full compatibility with LeRobot’s training pipeline. Check out thedocumentation.

https://huggingface.co/blog/lerobot-release-v050#codebase-a-modern-foundationCodebase: A Modern Foundation

This release modernizes the codebase:

• Python 3.12+: LeRobot now requires Python 3.12 as the minimum version, enabling modern syntax and better performance.
• Transformers v5: We’ve migrated to Hugging Face Transformers v5, staying current with the latest model ecosystem.
• 3rd-party policy plugins: Just like v0.4.0’s hardware plugin system, you can now register custom policies as installable packages —pip install lerobot\_policy\_mypolicyand use it with\-\-policy\.type=mypolicy. No core library changes needed. Learn how to do it by following thedocumentation.
• Remote Rerun visualization: Visualize your robot’s telemetry remotely using Rerun, with compressed image support for bandwidth-efficient streaming.
• Installation improvements: Addeduvinstallation instructions, clarified setup steps, and improved dependency management. Sequential install steps are now clearly documented.
• Documentation versioning: Docs are now versioned, so you can always find documentation matching your installed release.
• PyTorch version bump: Updated PyTorch version bounds to support NVIDIA Blackwell GPUs.

https://huggingface.co/blog/lerobot-release-v050#community–ecosystemCommunity & Ecosystem

• Modernized Discord: Updated the most vibrant community hub with a better channel organization.
• GitHub README, templates & automated labeling: A refreshed README, new issue and PR templates, contributing guidelines, and automatic labeling of tickets — making it easier for everyone to contribute.
• ICLR 2026 paper acceptance: The LeRobot paperhas been accepted to ICLR 2026!
• LeRobot Visualizer refresh: The visualization tool got a refresh with new dataset visualization badges and improved functionality.Check it out !
• LeRobot Annotation Studio: A HuggingFace Space designed to easily annotate every moment of your dataset with natural language subtasks.Check it out !

https://huggingface.co/blog/lerobot-release-v050#final-thoughtsFinal Thoughts

Beyond these headline features, v0.5.0 includes hundreds of bug fixes, documentation improvements, CI/CD enhancements, and quality-of-life improvements across the entire codebase. From better type checking to more robust test infrastructure, we’re investing in the foundations that make LeRobot reliable and maintainable as it scales.

We want to extend a hugethank you to everyone in the community— contributors, users, and collaborators alike — for helping LeRobot grow into what it is today. Every bug report, PR, and discussion makes this project better.

Stay tuned for more to come 🤗 Get startedhere! – The LeRobot team ❤️

There’s a big surprise coming just right around the corner, stay tuned! 👕