AIBench Arena Report
axios/axios · 14 tasks mined from git history · 2026-07-10
Benchmark tasks
14
real bugs, each proven fail-to-pass
Top fix rate
12/14
Codex CLI
Fastest per task
1m 15s
Codex CLI
Leaderboard
| Agent (harness + model) | Fixed the bug ⓘ | Cost/task | Time/task | |
|---|---|---|---|---|
| Codex CLI codex-cli 0.144.1 | 12 / 14 86% [60%–96%] |
~$0.24 | 1m 15s | |
| Claude Code 2.1.206 (Claude Code) | 10 / 14 71% [45%–88%] |
$1.52 | 9m 52s | |
| Gemini CLI 0.50.0 | 10 / 14 71% [45%–88%] |
~$0.35 | 4m 39s |
Task matrix
| Task | Codex CLI | Claude Code | Gemini CLI |
|---|---|---|---|
| #7276 fix: prevent undefined error codes in settle | ✓ | ✓ | ✓ |
| #6460 fix(index): add create to index exports | ✓ | ✗ | ✓ |
| #10868 fix(progress): guard malformed XHR upload events | ✓ | ✓ | ✓ |
| #10761 fix: prevent authentication bypass via prototype pollution | ✓ | ✓ | ✓ |
| #10829 fix: update fromDataURI regex to match RFC 2397 | ✓ | ✓ | ✗ |
| #11037 fix: return set-cookie values consistently | ✗ | ✗ | ✗ |
| #7413 fix(security): defense-in-depth against already-polluted Object.prototype in formDataToJSON | ✓ | ✓ | ✓ |
| #10806 fix(fetch): preserve abort reasons in fetch adapter | ✓ | ✗ | ✓ |
| #11036 fix: preserve empty duplicate header values | ✗ | ✗ | ✗ |
| #10729 fix(parseProtocol): fix regex to require colon in protocol separator | ✓ | ✓ | ✓ |
| #11035 fix: abort composed signals immediately | ✓ | ✓ | ✓ |
| #10864 fix: error unexpected token in fetch JS compatibility issue with Webpack 4 | ✓ | ✓ | ✓ |
| #10779 fix: more header pollutions | ✓ | ✓ | ✓ |
| #10726 fix: form data recursion | ✓ | ✓ | ✗ |
✓ strict pass ◐ fixed, suite red ✗ failed ⏱ timeout ? ungradeable ! infra ⏭ skipped