OmniMapBench: leaderboard
OmniMapBench (2026) tests visual reasoning over map documents with 2,096 manually annotated single-choice, multiple-choice and ordering questions over 1,603 maps in nine categories. The final score is strict exact-match accuracy in percent over all questions with chain-of-thought prompting; higher is better.
Metric: Final score: overall strict exact-match accuracy on manually annotated single-choice, multiple-choice and ordering questions over map documents, chain-of-thought prompting on all 2,096 questions (%); higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 25 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 75.03 | #94 |
| 2 | Qwen 3.5 397B A17B | 72.22 | #246 |
| 3 | Qwen 3.5 Plus | 70.8 | #232 |
| 4 | Qwen 3.5 27B | 68.62 | #359 |
| 5 | Qwen 3.5 122B A10B | 68.37 | #331 |
| 6 | Kimi K2.5 | 67.64 | #256 |
| 7 | Qwen 3.5 35B A3B | 64.75 | #482 |
| 8 | Qwen 3.5 Flash | 64.56 | #435 |
| 9 | Gemini 2.5 Pro | 61.4 | #266 |
| 10 | GPT-5 | 56.54 | #167 |
| 11 | GPT-5 Mini | 56.11 | #348 |
| 12 | Gemma 4 31B (IT) | 51.67 | #342 |
| 13 | GLM-4.5V | 49.33 | #744 |
| 14 | GPT-4.1 Mini | 45.27 | #779 |
| 15 | Claude Sonnet 4.5 | 45.04 | #250 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=omnimapbench · How It Works · Data refreshed daily, snapshot 2026-11-13.