Mercury Edit 2 achieves ultra-fast next-edit predictions in just 221 milliseconds using a diffusion-based language model approach.

Mercury Edit 2 delivers next-edit predictions with a latency of only 221ms, representing significant speed improvements

The model uses a diffusion LLM architecture, combining diffusion models with large language model capabilities

Developed by Inception Labs, the technology targets rapid code editing and prediction workflows

The approach appears to advance the state of real-time code completion and editing assistance tools