Cognition launches SWE-1.7 AI model, achieving frontier-level coding intelligence at lower cost through improved reinforcement learning; outperforms previous versions on benchmark tests, now available in Devin platform at 1000 TPS via Cerebras.

SWE-1.7: Frontier Intelligence at a Fraction of the Cost

Ben Pan*, Carlo Baronio*, Rohan Choudhury, Eric Lu, Ryan Kim, Deniz Birlikci, TC Qin, Sam Lee, Fermi Ma, Allen Liu, Yang Liu, Sampriti Panda, Jacob Teo, Ray Wang, Gary Chang, Steven Cao, Silas Alberti*equal contribution07.08.26Today, we’re launching SWE-1.7, the most capable model we’ve trained so far. It reaches frontier-level intelligence at a much lower cost, advancing the cost-performance Pareto curve.SWE-1.7 is the result of broad improvements across our RL pipeline: better infrastructure, ...