Inference is a distributed GPU cluster for LLM inference based on Solana, providing a fast, scalable, and token-based API for models such as DeepSeek V3 and Llama 3.3.