AI Harness Leaderboard
The definitive ranking of real AI model + harness combinations. Ranking AI model + harness combinations across coding, reasoning, and autonomy.
#1 Overall Score76.9Claude Code
| # | Harness | Score | |
|---|---|---|---|
| 1 | Claude Code | 76.9 | |
| 2 | Roo Code | 76.1 | |
| 3 | Devin | 74.7 | |
| 4 | Kilo Code | 74.3 | |
| 5 | Hermes Agent | 72.6 | |
| 6 | OpenClaw | 72.6 | |
| 7 | Cline | 72.1 | |
| 8 | Antigravity | 70.5 | |
| 9 | Augment Code | 70.0 | |
| 10 | Aider | 68.4 | |
| 11 | Amazon Q Developer | 64.7 | |
| 12 | Windsurf | 64.4 | |
| 13 | Cursor | 62.0 | |
| 14 | Copilot Workspace | 60.7 | |
| 15 | Codex CLI | 51.1 |
๐Open โ BYOK, any provider๐Semi-Open โ BYOK with constraints๐Closed โ fixed model set
Scores updated daily ยท Benchmark-derived with harness-specific adjustments ยท Not affiliated with any harness vendor