11h agoNVIDIA announces Nemotron 3 Ultra, a 950-billion-parameter SSM-MoE hybrid model claiming five times faster speedsArtificial Analysis ranks it ahead of Qwen3.5 on speed.SentimentSentimentPos61.2%Neg38.8%Many users praised NVIDIA's Nemotron 3 Ultra for its 5X faster inference, hybrid SSM-MoE design, and open-source elements that advance agentic workflows, while others dismissed the model as uncompetitive and criticized the focus on speed.30 comments with sentiment. View comments.