Benchmark
TextVQA
OCR benchmarks measure how well models read text in images and documents. Scores are percentages.
- Models measured
- 26
- Scores recorded
- 26
- Sources
- 2
- IQ weight
- 1.09
Score distribution
- 66.1%
- Mean
- 72%
- Median
- 21.3%
- Std dev
- 36.9%
- 10th percentile
- 84.4%
- 90th percentile
- 8.4%
- Min
This benchmark has a weight of 1.09 for chat models, measured over 25 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | Qwen2-VL-72B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 85.5% |
| 2 | Qwen2.5-VL 7B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 84.9% |
| 3 | Qwen2.5-Omni 7B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 84.4% |
| 4 | DeepSeek VL2· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 84.2% |
| 5 | InternVL2 26B· OpenVLM Leaderboard data | 82.5% |
| 6 | internvl3-8b· OpenVLM Leaderboard data | 82.1% |
| 7 | Nova Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 81.5% |
| 8 | Nova Lite· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 80.2% |
| 9 | MiniCPM-V 2.6· OpenVLM Leaderboard data | 78.3% |
| 10 | Phi 4 Multimodal Instruct· OpenVLM Leaderboard dataValues from 2 sources
| 76.9% |
| 11 | Pixtral 12B· OpenVLM Leaderboard data | 75.7% |
| 12 | InternVL2 4B· OpenVLM Leaderboard data | 74.7% |
| 13 | Phi-3.5 Vision Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 72% |
| 14 | GPT 4.1· OpenVLM Leaderboard data | 71% |
| 15 | Molmo 7B-D· OpenVLM Leaderboard data | 70.3% |
| 16 | Gemma 3 12B IT· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 67.7% |
| 17 | Gemma 3 27B IT· OpenVLM Leaderboard dataValues from 2 sources
| 66.7% |
| 18 | GPT-4.1 mini· OpenVLM Leaderboard data | 64.9% |
| 19 | Gemini 1.5 Flash· OpenVLM Leaderboard data | 61.6% |
| 20 | GPT-4.1 nano· OpenVLM Leaderboard data | 60.9% |
| 21 | Gemma 3 4B IT· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 57.8% |
| 22 | Gemma 3 4B· OpenVLM Leaderboard data | 57.7% |
| 23 | gemini-1.0-pro· OpenVLM Leaderboard data | 36.9% |
| 24 | kosmos-2· OpenVLM Leaderboard data | 14.7% |
| 25 | Qwen-VL Max· OpenVLM Leaderboard data | 8.6% |
| 26 | Qwen-VL Plus· OpenVLM Leaderboard data | 8.4% |
Data credits: Data by LLM Stats · OpenVLM Leaderboard (OpenCompass / VLMEvalKit), Apache-2.0 (licence)