NVIDIA releases Isaac Lab 3.0 in general availability for robot learning

NVIDIA announced the general availability of Isaac Lab 3.0, the open-source robot learning framework that is now the recommended baseline for new projects and migrations from version 2.x.

Version 3.0 introduces a factory-based architecture that lets developers choose different physics, rendering, and visualization backends without rewriting task logic. Key backends include Newton (with solvers such as MuJoCo-Warp, VBD, and MPM), OVPhysX, and the full Isaac Sim stack when needed.

What changed

Highlights include:

  • Kit-less execution with Newton and OV libraries, without needing the full Isaac Sim install for most workflows.
  • Warp-native simulation data with zero-copy access to both PyTorch and Warp.
  • Unified CLI commands via isaaclab for training and playback across multiple RL libraries (RSL-RL, RL-Games, skrl, Stable-Baselines3, TorchRL, and others).
  • Improved support for sensors, contact forces, deformables, particles, and solver coupling.
  • Policy export tools via LEAPP and guidance for transfer between PhysX and Newton.

The release updates dependencies to Isaac Sim 6.2, Python 3.12, PyTorch 2.11 with CUDA 13.0, NVIDIA Warp 1.17, and Newton 1.6.1. Interactive browser demos allow testing Newton physics without installation.

Why it matters

Isaac Lab is one of the main open tools for training robot policies in simulation before deploying them in the real world. Version 3.0 lowers the entry barrier by enabling lighter and more flexible workflows, which should accelerate experiments in dexterous manipulation, locomotion, and tasks involving deformable objects or liquids.

NVIDIA positions Isaac as part of the path toward the “ChatGPT moment” for robotics, combining foundation models, synthetic data, and scalable simulation.

NVIDIA’s official announcement and the GitHub release confirm general availability. The @NVIDIARobotics post on X highlighted the new modular architecture and multi-physics support.

Sources

Transparency: This content was created, edited, or reviewed with the assistance of artificial intelligence. Information was cross-checked with public posts on X and sources available on the internet. Consult the original sources to verify the full context.

By GeekikiBot