OmniMapBench - Multi-Step Reasoning: leaderboard
Level 3 of OmniMapBench (2026): strict exact-match accuracy in percent on the 853 multi-step reasoning questions about map documents, with chain-of-thought prompting. Higher is better.
Metric: Level 3 (multi-step reasoning, 853 questions): strict exact-match accuracy on manually annotated single-choice, multiple-choice and ordering questions over map documents, chain-of-thought prompting (%); 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) | 62.88 | #94 |
| 2 | Qwen 3.5 397B A17B | 60.37 | #246 |
| 3 | Qwen 3.5 Plus | 57.47 | #232 |
| 4 | Kimi K2.5 | 56.71 | #256 |
| 5 | Qwen 3.5 27B | 53.97 | #359 |
| 6 | Qwen 3.5 122B A10B | 53.54 | #331 |
| 7 | Qwen 3.5 35B A3B | 49.93 | #482 |
| 8 | Qwen 3.5 Flash | 48.62 | #435 |
| 9 | GPT-5 | 48.53 | #167 |
| 10 | Gemini 2.5 Pro | 46.31 | #266 |
| 11 | GPT-5 Mini | 45.13 | #348 |
| 12 | Gemma 4 31B (IT) | 40.68 | #342 |
| 13 | GPT-4.1 | 37.87 | #495 |
| 14 | GLM-4.5V | 35.8 | #744 |
| 15 | GPT-4.1 Mini | 35.64 | #779 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=omnimapbench-multi-step-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-13.