# SWE-bench

Real GitHub issues as coding-agent tasks.

SWE-bench measures whether models can patch real open-source repositories from issue descriptions. It is the standard reference for software engineering agent benchmarks.

Price: Open benchmark and evaluation harness; compute is yours.. Site: https://www.swebench.com/.

Also on the [web page](https://aibenchmarkindex.com/apps/swe-bench).

> **Check before you compare.** Benchmark methods and leaderboards change. Follow the cited primary source before using a result in a decision.