13 Models and 4 Agents on SWE Tasks: Go, Java, Python, Rust, TS

SWE-rebench: A Continuously Evolving and Decontaminated Benchmark for Software Engineering LLMs