๐ค Alibaba's Qwen3.8-27B open-weights model beats rivals 10x its size on key benchmarks
Released August 14, Qwen3.8-27B is a 27.8B-parameter multimodal model that accepts text, images, and video โ and runs on just 24GB of VRAM under Apache 2.0. It scored 61.7% on SWE-Bench Pro and outperformed Meta's Muse Glimmer 30B on all 8 shared benchmarks. Could this be the most capable open local AI of 2026?
โก 47 buzzing
691 views
๐ 0 subscribed
