Inference is a decentralized AI infrastructure platform providing low-cost, high-performance compute for model training and inference. It aggregates underutilized GPU resources from data centers into a distributed marketplace, enabling developers to deploy, fine-tune, and run large language models (LLMs) via serverless APIs. Supporting models like Llama 3.2 and Gemma 3, Inference reduces compute costs by up to 90% while maintaining enterprise-grade performance.