NVIDIA releases Nemotron-Labs-3-Puzzle-75B-A9B compressed MoE

The model is a compressed hybrid MoE with 75B total parameters and 9B active, derived from Nemotron-3-Super-120B-A12B via Iterative Puzzle. It delivers 2.03x server throughput at matched user throughput and is available in NVFP4 format for multi-GPU inference.