Groq

Maker of the LPU inference chip

Groq designs custom AI inference chips (LPUs) built for very low-latency token generation, and offers a cloud API for running open models at high speed.