Announcing NVIDIA Nemotron 3.5 Super VL on Baseten

NVIDIA’s Nemotron 3.5 Super VL is an open multimodal model now available through Baseten for applications requiring visual understanding, reasoning, and tool use across documents, images, user interfaces, and video. Built on a 120-billion-parameter mixture-of-experts architecture that activates 12 billion parameters per token, the model combines a vision encoder with hybrid Mamba-Transformer processing, long-context support, multi-token prediction, and NVFP4 cache quantization to improve inference efficiency. NVIDIA reports that Super VL also exceeds Nemotron 3 Super on text-focused benchmarks including GDPval, SciCode, and GPQA-diamond, indicating that multimodal capabilities do not diminish reasoning performance. Baseten supports inference through Dedicated and Model APIs, while its Loops platform enables managed supervised fine-tuning and reinforcement learning, allowing teams to customize, deploy, and serve trained model checkpoints within the same environment for use cases such as customer service, document intelligence, enterprise productivity, and web research.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.