Qwen 3.8 27B: excellent local model, but defaults to overthinking
Alibaba's Qwen 3.8 27B, an Apache 2 licensed 27B vision-capable LLM, ships with a default reasoning effort of "xhigh" that causes spectacular over-thinking. Simon Willison found it excellent on his 128GB M5 Max MacBook Pro and NVIDIA DGX Spark, but recommends raising the context limit to 262,144 tokens.
How this story unfolded
4 weeks · 3 reports · 109 community posts · 112 of 124 shown
- Aug 3
- Aug 11
- Aug 12
- Aug 13
- Aug 14
- Aug 15
- Aug 16
- Aug 17
- Aug 18
- Aug 19
- Aug 20
- Aug 21
- Aug 22
- Aug 23
- Aug 24
- Aug 28
Qwen by email
Get an email when Qwen has news
No news that day, no email.
More stories today
- AI band gets YouTube Official Artist Channel status
- Sony Music, Warner sue Anthropic over alleged IP theft
- AWS engineer shows robot answering untrained questions
- Bernie Sanders vows legislation to stop Flock AI surveillance
- Claude-built market-timing game: couch beats players 62%