Issue 2026-11-03 · Industry · 模型发布 · 应用
User says Qwen 3.8 27b used Blender to do math
A LocalLLaMA user said that when given a math-heavy physics problem, Qwen 3.8 27b found it could run Python via Blender MCP and wrote and executed code to speed up its math. The user called the behavior unprecedented for a local model and warned that models need explicit limits.
Read original ↗- Study probes Qwen's 'Wait' reasoning loops
- User mines 16 years of city council notes with local Qwen model
- Open-source research proposes new sparse attention for ultra-long context
- Alibaba Qwen Code Repo Reportedly Leaks qwen4-max and Other Model IDs
- Study questions LLMs' ability to read neural net weights