Model Architecture & Training MethodsInception Labs et al.
Inception Labs announces Mercury
Mercury announced — with diffusion-based parallel token generation, Mercury Coder Mini reaches 1,109 tok/s on an H100
Why it matters
Claims up to 10× the speed of speed-optimized frontier models. The first time a diffusion LM entered a commercial track
Source — Primary sources first; secondary reports are labelled “Press”
Paperarxiv.orghttps://arxiv.org/abs/2506.17298Added in Week 35, 2026 · See that week