#24260 merged Llama.cpp Arch Cohere-Moe Support Added

In three linesPR #24260 merged into llama.cpp adds Cohere-MoE architecture support. Users testing North Mini Code models, which are ~3GB smaller in Q8 quantization compared to Qwen 3.6 27B for local inference on homelabs.

Summary generated by Claude — human-verified