Benchmark Almanac

Nebius

We track 2 benchmarks that Nebius made or helped make.

0 in use 1 with public code 2 with a leaderboard here

Benchmarks

Name Kind Released Best score Status
SWE-rebench Agents fix fresh GitHub issues that an automated pipeline mines each month. Tests decide whether the issue is resolved. Fixing real repositories May 2025 64.5% System cards only
SWE-rebench V2 Agents repair real repository issues using executable tests. The authors compare models on 300 tasks across five languages. Harbor also ports 51 validated Clojure tasks. Fixing real repositories Feb 2026 25.2% Not in cards