Mercury

Mercury is a diffusion LLM (dLLM) that generates text in parallel, yielding a 5-10X speed-up for coding workflows.

Mercury is a diffusion LLM that makes AI feel instant. The model's breakthrough diffusion-based generation process delivers responses 5-10x faster than traditional LLMs. As a small model, with a quality comparable to GPT 4.1 Nano and Claude 3.5 Haiku (as measured by Artificial Analysis), Mercury is ideal for speed-sensitive tasks.

Mercury supports a 128K context length, tool calling, structured outputs - plus a rich set of tuning options to tailor deployments to your needs.