FrontierCode — leaderboard
Cognition benchmark for production-quality coding agents measuring whether maintainers would merge model PRs. It uses 150 maintainer-authored open source tasks with nested Extended, Main, and Diamond subsets scored by blockers and quality rubrics.
Metric: Main Score (self-reported). Source: benchmarklist.com. Status: saturation imminent. 16 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 5 | 38.8 |
| 2 | Claude Opus 4.8 | 34.3 |
| 3 | Claude Opus 4.8 (xHigh) | 34.3 |
| 4 | GPT-5.5 | 25.5 |
| 5 | GPT-5.5 (xHigh) | 25.5 |
| 6 | Claude Opus 4.7 | 23 |
| 7 | GPT-5.4 Mini | 17.8 |
| 8 | Gemini 3.1 Pro (Preview) | 16.7 |
| 9 | Claude Sonnet 4.6 | 15.1 |
| 10 | MiniMax-M2.7 | 6 |
| 11 | MiniMax-M2.5 | 5.3 |
| 12 | Gemini 3.1 Flash Lite (Preview) | 4.8 |
Interactive version: theaggregate.ai/benchmark?slug=frontiercode · How the rankings work · Data refreshed daily, snapshot 2026-07-22.