🚨 Moonshot AI researcher Xinyu Yang is already teasing Kimi K3.1😂

In my opinion from trying the model out the next upgrade needs to bring:

• An improvement in speed/inference • More token efficiency, especially on bigger tasks while reasoning • In general to be more reliable and close the gap between K3 and Fable/GPT-5.6

What about Kimi K3 would you improve specifically?