Agentic software-engineering benchmark built from complex real-world repositories and multi-file tasks.
- Launched
- Sep 19, 2025
- Models
- —
- Modality
- —
- Open source
- —
- API available
- —
- Releases
- 1
Releases
Sep 19, 2025SWE-Bench Pro releasedBenchmark result
Scale released a harder, contamination-resistant benchmark using real and commercial repositories to measure software agents on complex multi-file engineering tasks.

