OpenVLM OCR-VQA — leaderboard

OpenCompass OpenVLM OCR-VQA test-core results: question answering over book-cover images where models must read title, author, genre, and other visual text.

Metric: Accuracy (%). Source: huggingface.co. 118 models tracked.

Top models

#ModelScore
1Qwen 2 VL 72B73.7
2Pixtral-12B64.7
3Aquila-VL-2B63.3
4InternVL2-8B42.6
5InternVL3-8B38.9
6GPT-4.1 Nano36
7InternVL3-14B30.5
8Gemma 3 4B27.6
9GPT-4.1 Mini26.3
10GPT-4.1 (2025-04-14)18.7
11Gemini 1.5 Flash18.5
12Gemma 3 12B17.7
13Gemma 3 27B13.5
14Gemini 1.0 Pro12.3

Interactive version: theaggregate.ai/benchmark?slug=openvlm-ocr-vqa · How the rankings work · Data refreshed daily, snapshot 2026-07-22.