Thanks for the prompt ideas as I'm testing qwen 3.8 27b on various harnesses!
Qwen 3.8 27b ran for 30-40 minutes.
NGL I'm really impressed.
Qwen 3.8 27B runs locally in one click on Magnitude
On my Macbook Pro M4 Max it gets between 10-20 tok/s Great model if you have the memory bandwidth to run it quicky. If you don't, an MoE model like Qwen 3.6 35B A3B will be a lot faster. Hoping to see a small Qwen 3.8 MoE!
Oneshot qwen 3.8 27b q4 on mac m5 (8 mins) and the second one on rtx4090 (just under 3 mins) -- same model
Prompt borrowed from: (thanks!)
You have reached the end of the archive
All of qwen38