Architectural RedlineBench: leaderboard
Metric: Mean Net Issue Score (%). Source: redlinebench.benfeicht.com. 19 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-6 | 65.5 |
| 2 | Claude Opus 5 | 62.7 |
| 3 | Claude Fable 5.1 | 62.4 |
| 4 | GPT-5.6 Sol | 49.1 |
| 5 | Claude Fable 5 | 47.9 |
| 6 | GPT-5.5 | 45.1 |
| 7 | Claude Opus 4.6 | 41.8 |
| 8 | Claude Opus 4.8 | 41.2 |
| 9 | GPT-5.5 Pro | 41.2 |
| 10 | GPT-5.4 Pro (xHigh) | 40.9 |
| 11 | Claude Opus 4.7 | 36.4 |
| 12 | Claude Haiku 4.5 | 25.5 |
| 13 | Claude Sonnet 4.6 | 21.8 |
| 14 | Gemini 3 Flash | 12.7 |
| 15 | Grok 4.20 | 10.9 |
Interactive version: theaggregate.ai/benchmark?slug=architectural-redlinebench · How It Works · Data refreshed daily, snapshot 2026-09-19.