Yhn HealthFitness first seen 1 d ago, last 48 min ago, peak #9

Qwen3.8 runs at 255 tokens per second on one RTX 5090

Original: Running Qwen3.8 at 255 tok/s on a single 5090

Engineers are discussing a setup that runs the Qwen3.8 language model at 255 tokens per second on a single Nvidia RTX 5090 graphics card, with details published by Mesh Applied. Achieving that speed on one consumer GPU is notable for local inference, since it suggests useful LLM performance is now possible without expensive multi-GPU servers.

Why now: High inference speed on a single consumer GPU challenges assumptions about the hardware needed to run large language models locally.

QwenNvidia RTX 5090Mesh Applied

Open on hn →

Rank over time, top of the chart is #1. 17 snapshots from 22 h ago to 48 min ago.

Evidence

• Running Qwen3.8 at 255 tok/s on a single 5090 · Ir0n · 14

API: https://socialmediatrends-api.osmike.com/v1/trends/2164021