OpenVLM SEED-Bench 2 Plus - Maps: leaderboard
SEED-Bench-2-Plus Maps accuracy from the OpenCompass OpenVLM Leaderboard (VLMEvalKit): multiple-choice questions on text-rich maps, one of the benchmark's three image types (charts, maps and webs). Accuracy in percent, higher is better.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 211 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 (2025-04-14) | 64.1 | #257 |
| 2 | Qwen 2 VL 72B | 63.4 | #450 |
| 3 | InternVL3-38B | 63.3 | #413 |
| 4 | Claude 3.5 Sonnet | 63.2 | #362 |
| 5 | GPT-4.1 Mini | 63.1 | #363 |
| 6 | InternVL3-78B | 62.7 | #361 |
| 7 | Qwen 2.5 VL 7B | 61 | #641 |
| 8 | Qwen 2 VL 7B | 60.7 | #743 |
| 9 | InternVL3-14B | 60.1 | #506 |
| 10 | InternVL3-8B | 60 | #624 |
| 11 | Gemini 1.5 Pro | 59.7 | #454 |
| 12 | Gemini 1.5 Flash | 59.6 | #620 |
| 13 | GPT-4.1 Nano | 59.1 | #730 |
| 14 | Claude 3.7 Sonnet | 58.6 | #259 |
| 15 | Pixtral-12B | 58.6 | #805 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=openvlm-seed-bench-2-plus-maps · How It Works · Data refreshed daily, snapshot 2026-11-05.