Nemotron 3.5 Local Test: Runs on 24GB RAM, Lags in Coding but Shines in Tool Calling

curiousily_ · reddit · 2026-08-12

A developer locally tested the Q5 quantized version of NVIDIA's Nemotron 3.5 Lightning (30B-A3B) on an M5 Pro (48GB RAM), sharing specific performance metrics:

Original post →

More from coding & agent

coding & agent channel →