Nvidia just dropped Nemotron 3, a 31.6B‑param LLM that blends Mamba’s hybrid tech with only 3B active weights per step. Think faster inference, lower cost—plus a new Artificial Analysis Index benchmark. Curious? Dive in. #Nemotron3 #MambaHybrid #LatentMoE

🔗 aidailypost.com/news/nvidias...