Learn about this first
Cognition released the SWE-2 Flash coding model, saying it outperforms all U.S. open-weight models on leading benchmarks, including models 3 times its size. The model is based on thinkymachines' U.S. model Inkling-Small, with 276B total parameters and 12B active parameters. Cognition says it costs 5 times less and runs 3 times faster than SWE-2.
Until the end of this month, all Pro, Max, and Teams users on Devin Desktop and Devin CLI can use SWE-2 Flash for free. Cognition said it is building a program to securely share the weights with partners and researchers, with more details to be announced in the coming weeks.
AI-generated from reports · updated 12 days agoLatest developments
Devin users can use SWE-2 Flash for free this month
Event progress2
SWE-2 Flash available for free for a limited time
Cognition: Devin users can use SWE-2 Flash for free this monthCognition releases SWE-2 Flash coding model
Cognition Models / Devin Blog: Cognition releases SWE-2 Flash coding model, raising Terminal-Bench 4.0 from 0.9% to 20.0%