Nemotron 3.5 Lightning lands firmly on the Intelligence / Speed Pareto Frontier.
- Source
- x.com
- Date
Nemotron 3.5 Lightning lands firmly on the Intelligence / Speed Pareto Frontier. It dominates Holo 3.1 35B A3B, a post-trained community variant of Qwen 3.6 35B A3B by @hcompany_ai, which was previously the best model in this region of the Intelligence / Speed tradeoff space.

Context
A model reaches a speed-capability frontier when no tested alternative improves both measures at once. Alex Cheema's chart places Nemotron 3.5 Lightning on that boundary: it shows a local.ai Intelligence Index score of 62.9 and task time of 1 minute 52 seconds, versus 61.0 and 2 minutes 11 seconds for Holo 3.1 35B A3B. Comparison chart.
Lightning therefore improves both plotted measures over Holo in this comparison. Other plotted models still offer higher scores or shorter task times, so its frontier position represents one useful tradeoff among the tested configurations.
- postIs the Rush to Build New LLM Inference Engines Fragmenting the Ecosystem?
- postBig model. 2.4T params, 95B active. 4.89TB. Surprised they didn't release a 4-bi
post@Jason @Lons @eisokant @ape Correction: Should actually be closer to 50 tok/sec
postApple markets a four-Mac cluster for trillion-parameter local inference
Checking sign-in…
Loading comments…


