Inference.net (Inference R&D, Inc.) provides a platform for deploying, monitoring, tracing, evaluating, and fine-tuning open-source and custom AI models, including a marketplace-style spot market for underutilized data center GPU capacity. It targets AI-native engineering teams that want to run inference on models such as Kimi, GLM, MiniMax, and GPT-OSS, or train task-specific models on their own production traces. Key capabilities include an OpenAI-compatible gateway, LLM observability, continuous evaluation, and fine-tuning workflows, with SOC 2 Type II certification.