Mac mini 本地连跑 11 个开源权重模型，含 DeepSeek V4.1 Flash

Pratik Gandhi @pratikg原推发布于 2026.10.24

在 X 上打开原推

It will be a cold weekend outside but a hot one for my Mac mini. I'm going to be running 11 open-weight models NONSTOP! The challenge is giving them all the same piece of writing to work on. BUT there's a twist! I will save that for the results. The biggest is about 5x larger than my Mac's memory, so it will run off the SSD. Everything will stay on the machine locally. Here's the lineup: - DeepSeek V4.1 Flash - GLM 5.3 Flash - Qwen3.8 Flash Next (pruned) - Qwen3-Next 80B - Qwen3.6 35B - Ternary Bonsai 2 9B - GLM 4.7 Flash - Qwen3.8 27B - Gemma 4 26B - DeepSeek V2 Lite - Qwen3.5 0.8B I'll share the results as soon as it's done. Stay tuned.