NVIDIA releases Nemotron 3.5 Bolt E4B 4GB, a small open model for local agents

NVIDIA has released Nemotron 3.5 Bolt E4B 4GB, an open small language model. The model is post-trained from Gemma 4 E4B.

The release targets local voice assistants, game characters and robots. NVIDIA is aiming for about 4GB of VRAM on RTX and Jetson devices.

The announcement is brief and does not include benchmark results, licensing terms or download details, so those points are not covered here.

Developers building local voice or robotics agents now have an open model sized for NVIDIA RTX and Jetson hardware, which may reduce reliance on cloud inference.

Check the license and measure latency and accuracy on your own hardware before committing, since the announcement only gives a VRAM target.