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: Score (%). Source: cognition.ai. Status: years away from saturation. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.8 | 13.4 |
| 2 | GPT-5.5 | 6.3 |
| 3 | Claude Opus 4.7 | 5.2 |
| 4 | Gemini 3.1 Pro (Preview) | 4.7 |
| 5 | GPT-5.4 Mini | 4.6 |
| 6 | Kimi K2.6 | 3.8 |
| 7 | Claude Sonnet 4.6 | 3.5 |
| 8 | MiniMax-M2.7 | 2.4 |
| 9 | MiniMax-M2.5 | 1.1 |
| 10 | Kimi K2.5 | 1 |
| 11 | Gemini 3.1 Flash Lite | 0.7 |
Interactive version: theaggregate.ai/benchmark?slug=frontiercode · How It Works · Data refreshed daily, snapshot 2026-09-05.