OpenVLM SEED-Bench 2 Plus - Charts: leaderboard
SEED-Bench-2-Plus Charts accuracy from the OpenCompass OpenVLM Leaderboard (VLMEvalKit): multiple-choice questions on text-rich charts and diagrams, 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) | 71.7 | #257 |
| 2 | Gemini 1.5 Pro | 70.5 | #454 |
| 3 | GPT-4.1 Mini | 70.4 | #363 |
| 4 | Qwen 2 VL 72B | 70.1 | #450 |
| 5 | InternVL3-78B | 69.9 | #361 |
| 6 | Claude 3.5 Sonnet | 69.9 | #362 |
| 7 | InternVL3-38B | 69.4 | #413 |
| 8 | InternVL3-14B | 69.4 | #506 |
| 9 | Qwen 2.5 VL 7B | 68.4 | #641 |
| 10 | InternVL3-8B | 68 | #624 |
| 11 | Gemini 1.5 Flash | 68 | #620 |
| 12 | Claude 3.7 Sonnet | 67.7 | #259 |
| 13 | Gemma 3 27B | 65.8 | #606 |
| 14 | Moonshot v1 8K | 65.7 | #622 |
| 15 | Pixtral-12B | 65.4 | #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-charts · How It Works · Data refreshed daily, snapshot 2026-11-05.