Qwen 3.8 Max is one of the best models I've tested for running actual business workflows.
Not writing. Not answering questions. Running multi-step jobs inside my Agent OS. And one benchmark explains why.
Qwen 3.8 27B Uncensored 4-bit MLX Version running on M4 Pro 24 Gb.
I asked it a few uncensored question missile launch and virus designs and I am amazed tbh. (Only for educational purposes)
This guy is running qwen 3.8 27B locally on
5090 32 gb vram 115 t/s crazy speed
China's Qwen 3.8-27B just matched Claude Opus 4.6.
There's a free model that keeps up with the best AI from 6 months ago. And it runs on your laptop. Here's why this is wild: → It's tiny. Just 27B parameters → Fully open. Apache 2.0. Download it and it's yours forever →
You have reached the end of the archive
All of qwen38