SWE-bench Pro

DeepSWE: New Benchmark Sets Stricter Standards for AI Coding Agents