Nemotron 3.5 Lightning

Nemotron 3.5 Lightning is NVIDIA’s open-weight mixture-of-experts model, combining Mamba-2, expert layers, and select attention layers. It has roughly 30B total parameters with about 3B active per token, supports up to a 1M-token context window, and is built to run on a single 80GB GPU (H100 or A100) in BF16, or on smaller consumer hardware like an RTX 5090 using a 4-bit NVFP4 build.