NVIDIA releases Nemotron 3.5 Bolt E4B 4GB for local assistants and robots, with tool use and a 64K-token context window; the Gemma-derived model takes text input and produces text.

A real Reachy Mini robot in NVIDIA’s original Nemotron Bolt E4B demonstration; the language model handles conversation and tool use as part of a larger robot pipeline.

Published 23 Oct 2026 · 19:37 UTC

NVIDIA ↗ · Source dated 23 Oct 2026

NVIDIA releases Nemotron 3.5 Bolt E4B 4GB for local assistants and robots, with tool use and a 64K-token context window; the Gemma-derived model takes text input and produces text. · ournews.ai