Qwen3.8 27B draws community praise and local benchmarks
Qwen3.8 27B, a 27.3B parameter multimodal model with 262,144-token context, runs at ~14 tokens/s on a Mac Studio M3 Ultra via Ollama. Simon Willison calls it the most fun local model he's used, while community quantizations and GGUF builds proliferate.
How this story unfolded
4 weeks · 9 reports · 130 community posts · 139 of 150 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 26
- Aug 28
- Aug 29
Qwen by email
Get an email when Qwen has news
No news that day, no email.
More stories today
- Claude Code silently deletes local history older than 30 days
- Split sigmas speed up video reference in Stable Diffusion
- MiniMax tool turns anything into realistic human video
- MirroS' Code-as-World rewrites videos into executable MuJoCo programs
- PRAXIST research system boosts MLE-bench scores 44%