Kimi and GLM are some of the best open models available. They're also some of the hardest to serve efficiently. Check out this writeup to learn how we're making them faster and cheaper without losing accuracy. https://blog.cloudflare.com/smaller-fast…