Smaller specialized models can match or beat frontier generalists on the tasks they're trained for. Working with Applied Compute, we RL-trained SWE-check, a bug detection model that matches Opus 4.6 on our internal evals while running ~10x faster.
Today we're launching Cognition Japan, our first expansion into Asia, to partner with Japanese enterprises ready to transform how software gets built.
We're releasing SWE-1.6, our latest model optimized for both intelligence and model UX.
Devin can now break down large tasks and delegate them to a team of managed Devins, with each running in its own isolated VM in parallel.