Qwen-Next vs 27B for coding
A Reddit user compared Qwen-Next and 3.8 27B for coding on MTPLX, running q8 27B on an m5 max with 128GB RAM. They noted MTPLX's "optimized for speed" is a dynamic q4 with 8-bit attention and praised Alibaba for releasing the models for free.