ResearchTasksAgent task executionAgent task executionFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedQWENCLAWBENCHClaude Opus 4.659.5Overall Score18Jun 16, 2026SkillsBench 11 domainsHuman44.8Overall Score16Jun 8, 2026SkillsBench 1.0 (test)EvoSkills (Claude Opus 4.6)71.1Pass Rate (With Skills)8Apr 3, 2026AppWorld online modeSkillComposer-4B57.7Accuracy3Jun 5, 2026