Qwen3.8-27B tops Hugging Face trending, praised by Simon Willison

Alibaba's Qwen3.8-27B, an Apache 2.0 licensed 27B parameter vision-capable LLM, became the #1 trending model on Hugging Face. Simon Willison praised it as the most fun local model he's used, noting it defaults to 'xhigh' reasoning effort causing overthinking. Community quantizations and local benchmarks show strong performance, with one user reporting it beats Opus 4.8 on an RTX 5090.