Cerebras

Cerebras builds the world's largest AI chips — wafer-scale processors the size of a dinner plate — and uses them to deliver inference speeds up to 2,000+ tokens per second, making it the fastest publicly available AI inference platform. Its cloud service runs Llama and other open models at speeds that enable real-time applications previously impossible with GPU-based systems. Enterprises with ultra-low-latency requirements for voice AI, trading, and real-time analysis are its primary customers, and its inference API is available to developers.