SADU: leaderboard
Metric: Exact-match accuracy (%) over all questions of SADU (154 software architecture diagrams (behavioral, structural and ER) with 2,431 counting and retrieval questions, temperature 0); higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 70.18 |
| 2 | Gemini 2.5 Flash | 69.68 |
| 3 | Gemini 3.1 Flash Lite (Preview) | 66.31 |
| 4 | Claude Sonnet 4.5 | 56.36 |
| 5 | GPT-5 Nano | 55.45 |
| 6 | Claude Haiku 4.5 | 48.25 |
| 7 | Gemini 2.5 Flash Lite | 43.03 |
| 8 | Qwen 2.5 VL 7B | 38.58 |
| 9 | GPT-4o Mini | 17.77 |
Interactive version: theaggregate.ai/benchmark?slug=sadu · How It Works · Data refreshed daily, snapshot 2026-10-07.