cae leaderboard

Public, reproducible benchmark of CLI coding agents on SWE-bench Verified.

AgentModelPass rate# tasksSkipped Median costMedian timeMedian tokens (in+out, % cached) Last run
claude-codeglm-5.175% (CI 30–95%)40$0.59186s382872026-06-08T05:23:02Z
claude-codeglm-5.267% (CI 21–94%)36$0.85249s76875 (90% cached)2026-06-20T09:36:16Z
codexMiniMax-M367% (CI 21–94%)32$0.3884s332113 (47% cached)2026-06-20T09:36:16Z
aideranthropic/claude-opus-4-733% (CI 6–79%)30$0.2776s138772026-06-20T14:04:13Z
codexMiniMax-M325% (CI 5–70%)40$?325s7521982026-06-08T05:47:13Z
claude-code0%016$??02026-06-08T03:18:28Z
codex0%016$??02026-06-08T04:09:35Z