NVIDIA open-sources Nemotron-3.5-Super-VL: 120B weights, datasets, recipes and tech report
kuchaev · x · 2026-10-15
NVIDIA has released Nemotron-3.5-Super-VL (120B total / 12B active parameters plus a vision encoder) with a fully open stack:
- What's open: Base BF16 and post-trained BF16/NVFP4 weights, datasets, training recipes, cookbooks, and a full technical report
- Local deployment: The NVFP4 quantized variant reportedly runs well on a DGX Spark
- The model is live on Hugging Face and available via NIM API with vision input and thinking mode support
For developers studying post-training pipelines or running large open models locally, the weights-plus-data-plus-recipes bundle is a substantial release.
Related event: NVIDIA Open-Sources Nemotron 3.5 Super VL: 120B Multimodal Agent Model(28 posts)→
More from Infra
- Next infra frontier: self-evolving systems that adapt and self-verify — _shreya_s · 2026-10-16
- Cerebras pushes disaggregated inference: heterogeneous prefill pools feeding wafer-scale decode — ai · 2026-10-16
- d-Matrix GA's Corsair rackscale, demos first public Attention-FFN disaggregation at 845 tok/s — ai · 2026-10-16
- Actual's inference engine hits 10k tok/s on NVIDIA DGX Station, GB300 case study — natesiggard · 2026-10-16
- TSMC Q3 revenue hits $46.7B, up 41% YoY; Q4 guidance of $52-54B tops estimates — Beth_Kindig · 2026-10-16
- Neon Realtime turns any Postgres query into a live subscription, built inside the database — matei_zaharia · 2026-10-16