Skip to content
opencodex

Benchmarks

These are static snapshots of public leaderboards, refreshed manually — not live OpenCodex metering. Each board lists its source, capture date, and license note. Score-per-dollar rankings only appear on boards where every row carries a source-measured cost per task.

Boards are grouped by task domain:

  • Coding — DeepSWE, AA Coding Agent, FrontierCode, FrontierSWE, Program Bench, SWE Marathon
  • Frontend — Frontend Code Arena
  • Terminal — Terminal Bench 2.1
  • Security — Cybench
  • Intelligence — AA Intelligence Index