Cognition developed SWE-1.7, a model built from the open-source Kimi K2.7 Code base, and created an evaluation suite to measure trustworthiness by testing for propaganda output and consistent coding behavior across contexts. The results show SWE-1.7 performs comparably to leading U.S. frontier lab models despite starting from an open-source foundation, demonstrating that models derived from open-source bases can be trustworthy with sufficient care in development.