OpenVLM MTVQA: leaderboard
OpenCompass OpenVLM evaluation of MTVQA: multilingual text-centric visual question answering over images containing text in Arabic, German, French, Italian, Japanese, Korean, Russian, Thai, Vietnamese, and Chinese.
Metric: Accuracy (%). Source: arxiv.org. Source: huggingface.co. Saturation forecast: Estimated already saturated. 157 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 Mini | 36.8 | #355 |
| 2 | GPT-4.1 (2025-04-14) | 34.1 | #250 |
| 3 | InternVL3-78B | 32.5 | #352 |
| 4 | InternVL3-38B | 32.4 | #405 |
| 5 | Gemini 1.5 Pro | 32 | #445 |
| 6 | InternVL3-14B | 31.1 | #505 |
| 7 | Claude 3.7 Sonnet | 30.6 | #249 |
| 8 | Gemini 1.5 Flash | 30.5 | #618 |
| 9 | InternVL3-8B | 30.4 | #620 |
| 10 | Qwen 2 VL 72B | 29.6 | #460 |
| 11 | Qwen 2.5 VL 7B | 29 | #643 |
| 12 | Gemma 3 27B | 28.9 | #607 |
| 13 | GPT-4.1 Nano | 27.6 | #731 |
| 14 | Qwen 2 VL 7B | 25.8 | #750 |
| 15 | Gemma 3 12B | 23.8 | #671 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=openvlm-mtvqa · How It Works · Data refreshed daily, snapshot 2026-10-20.