Hands-on: NVIDIA Nemotron 3.5 Hits 250+ tokens/sec on 30B MoE
Arindam_1729 · x · 2026-08-11
A developer tested NVIDIA's newly released Nemotron 3.5 Lightning model. It is a 30B parameter MoE model (3B active parameters) designed specifically for high-volume, low-latency agentic tasks.
On the Nebius platform, actual inference speed reached an impressive 250-260 tokens/s. The tester noted that beyond its extreme speed, the model's reasoning performance is highly impressive, making it very promising for agentic workloads.
Related event: NVIDIA Unveils Nemotron 3.5 Lightning and NeMo Switchyard Router(55 posts)→
More from Models
- Prediction: frontier AI revenue growth could slow even as AI usage keeps rising — yi111 · 2026-10-16
- Meta teased 816B-param open-weight Muse Spark Open with 1M context, says meme thread — alexandr_wang · 2026-10-16
- Nolla Health Post-Trains Nemotron 3.5 VL: Visual Diagnosis Jumps from 35% to 66.8% — LoganGrasby · 2026-10-16
- Dev argues coding models now beat nearly all engineers in every branch of software — zack_overflow · 2026-10-16
- Prediction: Chinese frontier models could match GPT-6 by March 2027 — teortaxesTex · 2026-10-16
- Liquid AI's d1-ultra, #1 on Artificial Analysis Decision Index, lands on Vercel — JosephJacks_ · 2026-10-16