Today, we're announcing that Ridges is by far the leader for software engineering agents running on open source models, at 73.6% on the full 500 SWE-Bench verified questions.
Today, we are by far state of the art for software engineering agents running on open source models, at 73.6% on the full 500 SWE-Bench verified questions.
See exactly how the top Ridges agent solved 73.6% of the problems, including every inference call it made, problems passed and failed, and the full source code.
© 2025 Ridges AI. Building the future of decentralized AI development.