64 concurrent agents on one GB300: nemotron 120B hits ~7K tok/s aggregate
Sentdex · x · 2026-10-18
Sentdex shares a home agent-swarm experiment: running NVIDIA Nemotron 120B A12B on a single Asus DGX Station (GB300), he sustains 64 concurrent agents at 100 tok/s each, for nearly 7K tok/s aggregate. He jokes that Dario begged him in DMs not to do agent swarms at home, and quips that the agents did "NOT magically just start hacking the planet" — a jab at swarm-hype versus real capability.
More from coding & agent
- Shopify CEO Tobi Lütke: "We are now all verification engineers" — schwentker · 2026-10-18
- Shopify's Tobi Lütke: 'We are all verification engineers now' as agents become the third customer — schwentker · 2026-10-18
- Notion on an agent that opened a browser to edit by hand: "Models don't give up. And that's not always good" — schwentker · 2026-10-18
- Anthropic researcher on agent memory: "dreaming" aggregates memories from past sessions — schwentker · 2026-10-18
- Voice demo built with Jev draws "holy crap, that's the missing thing to get rid of keyboards" reaction — schwentker · 2026-10-18
- Agents reach llms.txt in only 32% of cases; 86% of those via site links — schwentker · 2026-10-18