Smaller, faster, safer: running Kimi and GLM at scale

Cloudflare's Workers AI runs Moonshot's Kimi K-series and Z.ai's GLM on GPUs in its data centers, close to users. The post details how quantization shrinks the large, long-context mixture-of-experts models so they run smaller, faster, and safer at scale.