NVIDIA releases Nemotron 3.5 Bolt, a 4GB open model for local agents

A 4GB open model tuned for multi-turn tool calling, with a reasoning toggle and published latency on Jetson and RTX, gives builders a concrete option for fully local voice, NPC and robot agents. Weights, data and training recipes are open.

• tool use — A model's ability to call external functions — run code, search the web, edit files — instead of only generating text.
• context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.

• articleStepFun Step 5 Preview: 600B open-weight agent model with NVFP4 recipes
• articleNVIDIA's PivotOPD trains agents to recover from early mistakes
• postNVIDIA Ships AIPerf for Load-Testing LLM Endpoints
• postNVIDIA and World Labs demo Atlas, a spatial world model, on real footage

• blogOpen From Data Center to Desktop: NVIDIA Expands Nemotron Family With Highest Efficiency Super VL and Bolt ModelsKari Briski
• postNVIDIA has just released the first Nemotron 3.5 model: Nemotron 3.5 Lightning,…Artificial Analysis
• blogNVIDIA Open Models Advance Europe’s Sovereign Intelligence Across IndustriesKari Briski

Checking sign-in…

Loading comments…