- Introduces NVIDIA Nemotron 3.5 Lightning, a 30B‑parameter MoE model (3B active) with up to 4x higher throughput and 30% faster task completion.
- Makes the model available via Amazon SageMaker JumpStart for one‑click deployment on AWS, with SDK integration and documentation links.
- Supports up to 1M token context via DFlash speculative decoding and full open‑trained licensing for post‑training and edge/on‑prem deployments.