Tweet by ritakozlov

August 3, 2026

4/ read about the optimizations we've made to run zai glm and moonshot kimi at scale (kv caching, compression & more) https://t.co/PYEXAydIn0