Comment by cogman10

I've not tried it out, but in the past I've been pretty disappointed in the performance of Kimi models. Deepseek, Qwen, and GLM all appear to perform much better even though Kimi often has higher benchmark scores. Ditto for Mimo.