Report
NVIDIA releases Nemotron 3.5 models, strengthening US open-weight AI options
Artificial Analysis says Super VL uses about 60% fewer output tokens per task than its predecessor.
TLDR
Artificial Analysis scores Nemotron 3.5 Super VL at 26.5 on its Intelligence Index, versus 13 for its predecessor, and says it uses about 60% fewer output tokens per task. It scores Bolt 74B-A9B 24GB at 21 and says both models accept vision and text input. The Super VL NVFP4 version scores within 0.1 points of its BF16 weights. Another user called Nemotron and Muse Spark a “big day for open weights in the US.”
Combined views
2.7K
2 Sources, first seen ago
