Currently, the best-performing models on SWE-Bench Pro score only 43.60% (claude-4-5-Sonnet) on the public dataset.

This market asks what the highest score will be on the public SWE-Bench Pro leaderboard by June 30, 2026, 23:59 UTC

SWE-Bench Pro contains 1,865 total tasks across 41 professional repositories, with the public set containing 731 instances. The benchmark contains complex, long-horizon tasks requiring edits across multiple files and repositories.

Resolution: Market resolves to the highest % Resolved score achieved by any model on the public SWE-Bench Pro leaderboard at https://scale.com/leaderboard/swe_bench_pro_public by June 30, 2026.

This market resolves to N/A if Scale AI discontinues the benchmark or stops measuring new models

SWE-Bench Pro (Public Dataset)

Explore the SEAL leaderboard with expert-driven LLM benchmarks and updated AI model leaderboards, ranking top models across coding, reasoning and more.