5 November 2026
Cognition releases SWE-2 Flash coding model, keeps weights private
First reported
Latent Space ran this on .
- Cognition, a coding-tools company, released SWE-2 Flash, a model built on Thinking Machines' Inkling-Small.
- The company says it outperforms American open-weight models, including some three times its size.
- It costs about 20% of what its larger SWE-2 model costs to run.
Our read
Only Latent Space, one of the 30 newsletters tracked, covered this release. It follows Google's faster coding version of Gemini 3.7 Flash, which our archive logged on 2026-08-18.
Earlier on this story
Written by Newskeryx from which newsletters covered this story and from our own archive.
How it was covered
Latent Spaceswyx & Alessio
SWE-2 Flash benchmarks, cost and speed, build timeline, weights policy, and pricing