Benchmark

SWE-bench Pro

Long-horizon software-engineering tasks solved in a real repository, scored on held-out test suites.

Score history →

Model scores

#ModelScore
1MiniMax M359%