What can a 4GB model do on a robot? Nemotron 3.5 Bolt E4B tested on Jetson
lmoroney · x · 2026-10-25
Laurence Moroney tested NVIDIA's new Nemotron 3.5 Bolt E4B — a Gemma 4 E4B base with Nemotron post-training (RLVR + RLHF), quantized to Q4KM and tuned for conversation plus multi-turn tool calls — running locally on a Reachy Mini robot with Cosmos handling vision, all on a Jetson Orin Nano 8GB. NVIDIA reports 4.2GB memory, 15 tokens/s, 64K context, and 35 languages. He highlights that the training recipes and datasets are published openly, and suggests a hands-on exercise: write ten test conversations for one small assistant task and run them before and after swapping models.
Related event: 4GB Nemotron Bolt E4B Runs Robots on Jetson(2 posts)→
More from Embodied
- AGI House stocks up on NVIDIA Jetson AGX Thor developer kits for upcoming build event — agihouse_org · 2026-10-25
- StackChan desk robot runs Nous Research hermes-agent fully locally with snappy voice — alexcovo_eth · 2026-10-25
- AppetiteCheck study: vagus stimulation cuts food intake 9.6% and eating speed 23.6% — plopesresearch · 2026-10-25
- EPFL's LogSync: a 4x2mm brain implant chip decoding cognitive states under 200 microwatts — petrusenko_max · 2026-10-25
- Samsung foldables at 3% after 7 years; Apple foldable predicted to hit 20% by 2028 — mitchdeg · 2026-10-25
- Figure CEO shows off three generations of its humanoid robots — adcock_brett · 2026-10-25