Nvidia releases Nemotron 3.5 Bolt E4B 4GB for local assistants

Nvidia says its new small open model builds on Gemma 4 E4B, supports multi-turn tool use and a 64K-token context, and is available through Hugging Face and Ollama. It’s designed for local assistants, interactive characters and robots. Across the benchmarks it evaluated, Nvidia claims roughly 40% higher average scores than its prior Nano 4B model.