EuraGovExam - History: leaderboard
Metric: Accuracy (%) on the history questions across the five regions; image-only: the model sees one scanned multiple-choice civil-service exam question in its original language and layout, with a fixed answer-format instruction and no OCR or tools; greedy decoding, one run; answers not in the required final-line format count as wrong; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 28 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 87.5 | #145 |
| 2 | GPT-5 | 86.33 | #91 |
| 3 | O3 | 86.33 | #121 |
| 4 | Gemini 3 Pro (Preview) | 74.22 | #64 |
| 5 | Gemini 3 Flash (Preview) | 71.09 | #78 |
| 6 | O4 Mini | 67.97 | #172 |
| 7 | Gemini 2.5 Flash | 66.02 | #237 |
| 8 | Claude Sonnet 4 | 66.02 | #194 |
| 9 | GPT-5.2 | 61.33 | #105 |
| 10 | GPT-5 Nano | 50 | #415 |
| 11 | GPT-4.1 | 48.83 | #240 |
| 12 | GPT-4.1 Mini | 39.84 | #346 |
| 13 | GPT-4o | 35.55 | #333 |
| 14 | Qwen 2.5 VL 7B Instruct | 26.95 | #643 |
| 15 | Qwen 2 VL 7B Instruct | 25.39 | #816 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=euragovexam-history · How It Works · Data refreshed daily, snapshot 2026-10-11.