NVIDIA releases Nemotron-3.5-Bolt-E4B small model

NVIDIA's new small model is post-trained from Gemma-4-E4B-it for low-memory, low-latency local voice assistants and edge devices, improving multi-turn tool use, instruction following and grounding. It is cleared for commercial and non-commercial use, with GGUF quantizations and a BF16 version available.

• Nvidia CEO demos Microduck robots, preorders exceed 20,0002026-10-23
• Train Microduck to walk in Isaac Lab 3.0 and deploy to real robot2026-10-21
• NVIDIA releases Nemotron-Labs-Dory1.0-4B model2026-10-17
• NVIDIA Nemotron Bolt 3.5 74B releases Oct. 152026-10-14
• Opinion: When AI agents evade human supervision2026-10-14