SWE-bench is saturated, so Cognition built FrontierCode
Cognition built FrontierCode eval after SWE-bench saturated, focusing on code mergeability
“okay, this code is technically correct. But would you actually merge it? Would you feel happy? Would this improve the quality of your code base?”
Cognition, maker of Devin, released FrontierCode after finding SWE-bench fully saturated by frontier models. The new benchmark targets 'mergeability' — whether code is stylistically sound, maintainable, and gracefully handles future changes — a quality dimension existing evals miss. Fable 5 has already pushed scores on the hardest FrontierCode Diamond subset from the teens to the 30s, signaling rapid benchmark pressure.