Nemotron 3 Ultra: Open Moe Hybrid Mamba-Transformer for Agentic Reasoning [pdf] Posted by victormustar on Thursday June 04, 2026 @09:06AM from the research.nvidia.com dept. https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3-Ultra-Technical-Report.pdf ( 2 comments | 14 points | on HN )
It is significantly bigger than Qwen for the same level of intelligence, but I think the key strength was inference speed.