Versions

2 versions, plus 14 variants and configurations · Metric: Resolved Tasks

Versions of SWE-bench
VersionReleasedModelsResults
SWE-bench Verified
9 variants of Verified
Variants of SWE-bench Verified
VariantConfigurationReleasedModelsResults
SWE-bench Verified — bash-only harnessbash-only harnessJuly 15, 202622
SWE-bench Verified — GitHub Copilot production harnessGitHub Copilot production harnessJune 2, 202611
SWE-bench Verified — Muse Spark corrected testsMuse Spark corrected testsApril 8, 202611
SWE-bench Verified — Droid harnessDroid harnessFebruary 12, 202611
SWE-bench Verified — OpenCode harnessOpenCode harnessFebruary 12, 202611
SWE-bench Verified — mini-swe-agentmini-swe-agent harnessJuly 12, 20253036
SWE-bench Verified — Epoch AI harnessEpoch AI harnessAugust 13, 20242628
SWE-bench Verified — Mercor mini-swe-agent, 1000 stepsMercor mini-swe-agent, 1000 steps, 3 hoursAugust 13, 20243435
SWE-bench Verified — Vals AI mini-swe-agentmini-swe-agent harnessAugust 13, 20245959
August 13, 20244551
SWE-bench MultilingualLatest
5 variants of Multilingual
Variants of SWE-bench Multilingual
VariantConfigurationReleasedModelsResults
SWE-bench Multilingual — Cursor standard harnessCursor standard harnessJune 25, 202611
SWE-bench Multilingual — Cursor strict harnessCursor strict harnessJune 25, 202611
SWE-bench Multilingual — GitHub Copilot production harnessGitHub Copilot production harnessJune 2, 202611
SWE-bench Multilingual — mini-swe-agentmini-swe-agent harnessFebruary 24, 20261212
SWE-bench Multilingual — Mercor mini-swe-agent, 1000 stepsMercor mini-swe-agent, 1000 steps, 3 hoursMay 6, 20253434
May 6, 20253135

Recently reported results: SWE-bench Multilingual

Latest version: SWE-bench Multilingual