NVIDIA releases Nemotron 3.5 Bolt E4B 4GB for local conversations

NVIDIA applied Nemotron post-training to Gemma 4 E4B and designed the open model to use roughly 4GB of memory on its target workloads. It supports a 64K context, 35 languages, multi-turn tool use, and control over reasoning. NVIDIA reports about 40% better average quality than Nemotron 3 Nano 4B on its main conversational agent benchmarks.

Visit original source