GSO: Software Optimization Benchmark for SWE-Agents
GSO evaluates AI coding agents on 102 challenging real-world software performance optimization tasks across 10 codebases in 5 languages, measuring whether an agent's patch matches expert-developer speedups while remaining correct.
What this benchmark measures
GSO evaluates AI coding agents on 102 challenging real-world software performance optimization tasks across 10 codebases in 5 languages, measuring whether an agent's patch matches expert-developer speedups while remaining correct.
Rows on this page are sourced from public benchmark artifacts, leaderboard exports, or source-linked model reports. Each row keeps benchmark version, source model name, and available run details attached to the score.
The metric shown here is Opt@1. It should be interpreted within GSO: Software Optimization Benchmark for SWE-Agents, not compared as part of a site-wide ranking.
Frequently asked
What is GSO: Software Optimization Benchmark for SWE-Agents?
GSO evaluates AI coding agents on 102 challenging real-world software performance optimization tasks across 10 codebases in 5 languages, measuring whether an agent's patch matches expert-developer speedups while remaining correct. It is a coding benchmark measured by Opt@1.
What does Opt@1 mean on GSO: Software Optimization Benchmark for SWE-Agents?
GSO: Software Optimization Benchmark for SWE-Agents reports Opt@1 (%); higher is better. Scores are shown only within GSO: Software Optimization Benchmark for SWE-Agents and are never averaged with other benchmarks.
What is the top reported GSO: Software Optimization Benchmark for SWE-Agents score?
Claude Opus 4.7 has the top reported score on GSO: Software Optimization Benchmark for SWE-Agents: 44.12% (Opt@1).
Why do GSO: Software Optimization Benchmark for SWE-Agents scores differ across runs?
Harness, scaffold, reasoning effort, and prompt setup change results, so two runs of the same model can differ. evals.report keeps each score with its run context so the differences stay visible.
Does evals.report rank models across benchmarks?
No. GSO: Software Optimization Benchmark for SWE-Agents scores are shown within their own metric; evals.report never combines benchmarks into a composite ranking or a single "best model".