Inside Nemotron 3.5 Super VL: hybrid Mamba-Transformer MoE with C-RADIOv4 vision encoder
vllm_project · x · 2026-10-15
The vLLM team details Nemotron 3.5 Super VL's architecture: a hybrid Mamba–Transformer MoE paired with a C-RADIOv4 vision encoder. vLLM supports the model's multi-token prediction (MTP) path for speculative decoding, alongside continuous batching and prefix caching.
The model has 120B total / 12B active parameters and a 1M-token context window, bringing text, image, document and video understanding with reasoning and tool use to an OpenAI-compatible serving endpoint.
Related event: vLLM ships Day-0 support for Nemotron 3.5 Super VL(4 posts)→
More from Infra
- Jensen Huang joins BG2 podcast: does AI revenue justify capex, '27 bottlenecks — elonmusk · 2026-10-16
- Mistral's on-device infra rack revealed: DGX Spark, Mac Studio and more — gregmushen · 2026-10-16
- NVIDIA and Sweden's Wallenberg launch Sferical One, a shared AI supercomputer — nvidia · 2026-10-16
- Next infra frontier: self-evolving systems that adapt and self-verify — _shreya_s · 2026-10-16
- Xiaomi's MiMo RL dashboard reveals $3.47M spent on just 30 training steps — bycloud · 2026-10-16
- Cerebras pushes disaggregated inference: heterogeneous prefill pools feeding wafer-scale decode — ai · 2026-10-16