NVIDIA Isaac Sim and Isaac Lab

What the evidence says

1. Company claim

   NVIDIA describes Isaac Sim as a simulation platform built on NVIDIA Omniverse, "designed to develop, test, train, and deploy AI-powered robots in realistic virtual environments", with GPU-accelerated physics, RTX-rendered sensors, synthetic data generation, reinforcement learning and ROS integration. Isaac Lab is its GPU-accelerated framework "built for reinforcement learning, imitation learning, and motion planning."

   Observed Oct 18, 2026 · NVIDIA (company source)
2. Documented by company source

   The Isaac Sim code is licensed under Apache 2.0, but NVIDIA states that "Building or using the software requires additional components licenced under other terms", including the Omniverse Kit SDK and 3D assets. Those components are covered by NVIDIA's Isaac Sim Additional Software and Materials License (version of June 9, 2025): a limited, revocable licence for systems with NVIDIA GPUs, for use with Isaac Sim or Isaac Lab, that bars redistribution and modification. Isaac Lab is licensed under BSD-3, with its isaaclab_mimic extension under Apache 2.0.

   Observed Oct 18, 2026 · NVIDIA (company source), NVIDIA Isaac Sim Additional Software and Materials License (company source), isaac-sim/IsaacLab (company source)
3. Documented by company source

   Isaac Sim 6.2.0 was released in October 2026 for Linux and Windows. Isaac Lab's latest stable release is 2.3.2 (January 29, 2026); version 3.0.0 has been in early access since September 16, 2026.

   Observed Oct 18, 2026 · NVIDIA (company source), isaac-sim/IsaacLab (company source)
4. Documented by company source

   NVIDIA lists a GeForce RTX 4080 with 16 GB of memory as the minimum GPU, an RTX 5080 as recommended and an RTX PRO 6000 Blackwell as best, and states: "GPUs without RT Cores (A100, H100) are not supported."

   Observed Oct 18, 2026 · NVIDIA (company source)
5. Customer-confirmed

   Agility Robotics wrote on its own site in October 2024 that it began using Isaac Lab early the previous year to train reinforcement-learning control policies for its Digit humanoid, and that GPU-accelerated simulation let it train far faster than CPU simulation. It also uses Isaac Sim to "explore the impact of different physical characteristics" for future versions of Digit. Agility does not describe the policies as a production deployment.

   Observed Oct 18, 2026 · Agility Robotics (named customer)

Price

• DownloadFreeFree. Requires NVIDIA-licensed components and an NVIDIA RTX-class GPU (at least a GeForce RTX 4080 with 16 GB). Checked Oct 18, 2026. NVIDIA (company source)