← All models

Nemotron 3.5 Super VL 120B A12B BF16

NVIDIA · 2026-10-15 · 124.5B parameters

NVIDIA's BF16 release of Nemotron 3.5 Super VL, a hybrid Mamba-2, attention and mixture-of-experts vision-language model (about 120B total and 12B active parameters) with a C-RADIO vision encoder, text, image and video input, up to 1M-token context and optional thinking, explicitly identified by LiveBench 2.0. The parameter count is the checkpoint's safetensors total; BF16 describes the distributed weights, not the training precision.

Benchmark scores