Observed arrival · 2026-10-05

Local AI Bench: a leaderboard that keeps the failures

local-ai-bench.org Visit website

Editorial interest 78/100 Selection signal · not a rating of the site

A public benchmark tracks Qwen3.8-Flash-Next and GLM-5.3-Flash runs on one Ryzen 9 5950X system with three RTX 3090s.

Landing page captured for the 2026-10-05 issue.

For: People comparing local model inference configurations
Worth noticing: Qualification combines speed with stability; the Flash-Next kit band also sets top-1 agreement and KL thresholds.

Field notes

The run table exposes configuration keys, stages, timestamps, and metrics such as decode speed, prefill speed, tool calls, and load duration. The GLM page distinguishes single-card results from three-card experiments, and labels below-threshold runs as documented experiments rather than registry submissions. The homepage shows no qualified eligible configuration for either model at the time represented in its snapshot.

Observed signals

Read the marks

Editorial observations of this landing page, not a rating.

○OpenPublic substance visible

✦PrettyNotable craft visible

●ProPolished or operationally mature

◎NicheUnusually specific use

One card from the complete issue

Still the Old Commit

307,787 arrived 1,000 judged 978 catalogued Enter the complete issue