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