ArchBench Traceability Link Recovery: leaderboard
Software-architecture benchmark for recovering traceability links between architecture documentation, design decisions, and implementation artifacts.
Metric: Weighted average F1 (%). Source: github.com. Saturation forecast: Estimated already saturated. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o (2024-08-06) | 86 | #336 |
| 2 | GPT-4 Turbo | 86 | #404 |
| 3 | GPT-4 (0613) | 86 | #491 |
| 4 | GPT-4o Mini (2024-07-18) | 85 | #558 |
| 5 | GPT-3.5 Turbo (0125) | 85 | #984 |
| 6 | Llama 3.1 8B Instruct | 68 | #1035 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=archbench-traceability-link-recovery · How It Works · Data refreshed daily, snapshot 2026-10-20.