NVIDIA

NVIDIA's diffusion-based 6-DOF grasp generator: give it a segmented object point cloud or mesh, get back ranked gripper poses with confidence scores. Checkpoints for Franka Panda, Robotiq 2F-140 and a 30 mm suction cup. Research use only.

GPU-accelerated robotics simulator on Omniverse/OpenUSD: PhysX 5 physics, RTX-accurate sensor simulation and Replicator synthetic-data generation — the reference platform for sim-to-real, digital twins and robot-learning pipelines (open source since 2025).

NVIDIA's CUDA-accelerated robot library: collision-free motion generation in tens of milliseconds via massively parallel trajectory optimization, GPU IK and MPC — the engine behind Isaac ROS cuMotion for reactive industrial manipulation.

NVIDIA's unified 6D object pose estimation and tracking model — model-based (CAD) or model-free (reference images), no fine-tuning for novel objects. #1 on the BOP leaderboard at release. Research use only.

Compact edge-AI module delivering up to 157 TOPS with an 8-core Arm CPU and 1024-core Ampere GPU, 10–40 W, for autonomous robots and rich perception.

Entry edge-AI module delivering up to 67 TOPS with a 6-core Arm CPU and 1024-core Ampere GPU in a 7–25 W envelope, for compact robots and edge vision.

Blackwell-based physical-AI module delivering up to 2070 FP4 TFLOPS with a 14-core Arm Neoverse CPU and 128 GB memory, 40–130 W, for humanoids and physical AI.

Server-class edge-AI module delivering up to 275 TOPS with a 12-core Arm CPU and 2048-core Ampere GPU, 15–60 W, for autonomous machines and humanoids.