SADU - Counting: leaderboard
Metric: Exact-match accuracy (%) on the counting questions of SADU across all diagram types (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) | 76 |
| 2 | Gemini 2.5 Flash | 72 |
| 3 | Gemini 3.1 Flash Lite (Preview) | 61 |
| 4 | Claude Sonnet 4.5 | 51 |
| 5 | GPT-5 Nano | 50 |
| 6 | Claude Haiku 4.5 | 44 |
| 7 | Gemini 2.5 Flash Lite | 32 |
| 8 | Qwen 2.5 VL 7B | 31 |
| 9 | GPT-4o Mini | 16 |
Interactive version: theaggregate.ai/benchmark?slug=sadu-counting · How It Works · Data refreshed daily, snapshot 2026-10-07.