# 99.5k tps on a B200

> With llama-3.1 8b

Published: 2025/11/04 • Author: Surya Dantuluri • Canonical: https://sdan.net/projects/hundredtps

Hit 99.5k tokens per second on a single NVIDIA B200 running Llama 3.1 8B. A Modal notebook demonstrating near-theoretical-max throughput with optimized inference settings.