00:00
2026-10-11
runagentrun.co.uk
large-language-models
Test your Qwen 3.8 27B quants
A Kelcode benchmark of three serving variants of Qwen3.8-Flash-Next against a Qwen3.8-27B control found that agentic tool-use tests separate the recipes while standard multiple-choice tests flatten th…