Mercury

Mercury is Inception Labs' family of diffusion large language models (dLLMs) delivered via API, featuring Mercury 2 (a fast reasoning LLM with 128K context, tool use, and structured output) and Mercury Edit 2 (a coding-focused model for latency-sensitive autocomplete and next-edit workflows), both designed for enterprise-grade production deployment with OpenAI-compatible APIs and support for fine-tuning and private deployments.