Nvidia Nemotron 3.5 Lightning: First Open-Source Model Post-Huang Letter

Nvidia releases Nemotron 3.5 Lightning, a 30B MoE (3B active) open-weight model with Mamba-2, 1M context, and speculative decoding. Runs on a single GPU, free to modify. Includes NeMo Switchyard cost optimizer for FinOps. Challenges hardware-only narrative and expands edge AI possibilities.