Nvidia Releases Nemotron 3.5 Super VL and Edge Model Nemotron 3.5 Bolt 74B-A9B

The pairing of a vision-language model with a 24GB-capable edge variant points to Nvidia pushing multimodal capability toward local and on-device deployment rather than cloud-only inference. That matters for builders weighing hardware constraints against model capability, and for the broader trend of shrinking capable models onto consumer or embedded hardware. The absence of pricing, availability timing and benchmarks leaves the practical question open: whether the edge model's footprint translates into usable performance, and whether Nvidia positions these as open releases or gated products.