SADU - Behavioral Diagrams: leaderboard

SADU (2026) tests vision-language models on counting and retrieval questions about behavioral, structural and entity-relationship software architecture diagrams. This board is exact-match accuracy in percent on questions about behavioral diagrams; higher is better.

Metric: Exact-match accuracy (%) on the 810 behavioral-diagram 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 February 2027. 11 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Flash (Preview)63.58#156
2Gemini 2.5 Flash62.1#466
3Gemini 3.1 Flash Lite (Preview)59.51#419
4Claude Sonnet 4.553.83#267
5GPT-5 Nano48.64#912
6Claude Haiku 4.548.27#587
7Qwen 2.5 VL 32B44.44#944
8Gemini 2.5 Flash Lite42.59#972
9Qwen 2.5 VL 7B34.2#1878
10Qwen 2.5 VL 3B29.26#2664
11GPT-4o Mini19.14#1680

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=sadu-behavioral-diagrams · How It Works · Data refreshed daily, snapshot 2026-11-22.