“The wall clock time for Kimi K3's API is significantly higher than for OpenAI's GPT models, partly due to high demand and servers being located in China.”