DocVQA
Seed1.5-VL 96.9% leads since May 2025 Models answer 50,000 questions about 12,000 scanned document images by reading their text and layout. Scoring is ANLS, a string similarity to reference answers.
Test answers stay on the evaluation server, and top models now pass the 94.36 human score in the paper.
- Made by
- CVC, Universitat Autònoma de Barcelona and IIIT Hyderabad
- Released
- July 2020
- Kind
- Documents, charts and text in images
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Short answer
- Tasks
- All public
- Strongest model tested
-
DeepSeek V4.1 Flash #34 on the capability index - Last in a launch post
- September 2026
- Last model added
- September 2026
In the authors’ words
We present a new dataset for Visual Question Answering (VQA) on document images called DocVQA. The dataset consists of 50,000 questions defined on 12,000+ document images. Detailed analysis of the dataset in comparison with similar datasets for VQA and reading comprehension is presented. We report several baseline results by adopting existing VQA and reading comprehension models. Although the existing models perform reasonably well on certain types of questions, there is large performance gap compared to human performance (94.36% accuracy). The models need to improve specifically on questions where understanding structure of the document is crucial. The dataset, code and leaderboard are available at docvqa.org
36
Models scored in lab cards
30
Launch posts list it, from 15 labs
23
System cards report it, from 11 labs
Seed1.5-VL leads with 96.9% among models released by September 2026
Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.
The newest model with a score here came out in September 2026, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- Claude Fable 5.2
-
Gemini 4 Carbon - Claude Opus 5.5
- GPT-6.1 Sol
- and 16 more
- 96.9 Seed1.5-VL
- 96.5
Qwen3-VL-235B-A22B - 96.4 Qwen2.5-VL-72B
- 95.9
Command A Vision - 95.6
Nemotron 3 Nano Omni 30B-A3B - 95.6 DeepSeek V4.1 Flash
- 95.6 DeepSeek-V4.1-Flash Base
- 95.5
ERNIE 5.0 - 95.3
Mistral Medium 3 - 95.2 Claude 3.5 Sonnet
- 95.2 Qwen2.5-Omni-7B
- 94.9
InternVL3.5-241B-A28B - 94.9 Mistral Small 3.2
- 94.4
Llama 4 Maverick - 94.4 Nemotron Nano 12B v2 VL
- 94.2 Claude 3.5 Sonnet (October 2024)
- 94.2 Cosmos3-Nano
- 94.1 Mistral Small 3.1
- 93.6 Grok-2
- 93.6 ERNIE-4.5-VL-28B-A3B-Thinking
- 93.2
Phi-4-multimodal - 93.1 Gemini 1.5 Pro
- 92.8 GPT-4o
- 92.1 North Micro Vision Instruct
- 91.8
LongCat-Flash-Omni - 91.2 Llama-3.1-Nemotron-Nano-VL-8B-V1
- 91.1 LFM2.5-VL-3B
- 90.9 Gemini 1.0 Ultra
- 89.5 Claude 3 Sonnet
- 89.3 Claude 3 Opus
- 88.8 Claude 3 Haiku
- 87.1 Gemma 3 12B
- 86.8 Cosmos3-Edge
- 86.6 Gemma 3 27B
- 86.5 Gemini 1.5 Pro
- 75.8 Gemma 3 4B
- 1 Seed1.5-VL 96.9%
- 2 Qwen3-VL-235B-A22B 96.5%
- 3 Qwen2.5-VL-72B 96.4%
- 4 Command A Vision 95.9%
- 5 Nemotron 3 Nano Omni 30B-A3B 95.6%
- 6 DeepSeek V4.1 Flash 95.6%
- 7 DeepSeek-V4.1-Flash Base 95.6%
- 8 ERNIE 5.0 95.5%
- 9 Mistral Medium 3 95.3%
- 10 Claude 3.5 Sonnet 95.2%
- 11 Qwen2.5-Omni-7B 95.2%
- 12 InternVL3.5-241B-A28B 94.9%
- 13 Mistral Small 3.2 94.9%
- 14 Llama 4 Maverick 94.4%
- 15 Nemotron Nano 12B v2 VL 94.4%
- 16 Claude 3.5 Sonnet (October 2024) 94.2%
- 17 Cosmos3-Nano 94.2%
- 18 Mistral Small 3.1 94.1%
- 19 Grok-2 Aug 2024 93.6%
- 20 ERNIE-4.5-VL-28B-A3B-Thinking 93.6%
- 21 Phi-4-multimodal 93.2%
- 22 Gemini 1.5 Pro May 2024 93.1%
- 23 GPT-4o May 2024 92.8%
- 24 North Micro Vision Instruct 92.1%
- 25 LongCat-Flash-Omni 91.8%
- 26 Llama-3.1-Nemotron-Nano-VL-8B-V1 91.2%
- 27 LFM2.5-VL-3B 91.1%
- 28 Gemini 1.0 Ultra 90.9%
- 29 Claude 3 Sonnet 89.5%
- 30 Claude 3 Opus 89.3%
- 31 Claude 3 Haiku 88.8%
- 32 Gemma 3 12B 87.1%
- 33 Cosmos3-Edge 86.8%
- 34 Gemma 3 27B 86.6%
- 35 Gemini 1.5 Pro Feb 2024 86.5%
- 36 Gemma 3 4B 75.8%
Source: Labs’ model cards
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Seed1.5-VL | 96.9% | May 2025 | Lab model card | |
| 2 | Qwen3-VL-235B-A22B | 96.5% | Sep 2025 | Lab model card | |
| 3 | Qwen2.5-VL-72B | 96.4% | Jan 2025 | Lab model card | |
| 4 | Command A Vision | 95.9% | Jul 2025 | Lab model card | |
| 5 | Nemotron 3 Nano Omni 30B-A3B | 95.6% | Apr 2026 | Lab model card | |
| 6 | DeepSeek V4.1 Flash | 95.6% | Sep 2026 | Lab model card | |
| 7 | DeepSeek-V4.1-Flash Base | 95.6% | Sep 2026 | Lab model card | |
| 8 | ERNIE 5.0 | 95.5% | Feb 2026 | Lab model card | |
| 9 | Mistral Medium 3 | 95.3% | May 2025 | Lab model card | |
| 10 | Claude 3.5 Sonnet | 95.2% | Jun 2024 | Lab model card | |
| 11 | Qwen2.5-Omni-7B | 95.2% | Mar 2025 | Lab model card | |
| 12 | InternVL3.5-241B-A28B | 94.9% | Aug 2025 | Lab model card | |
| 13 | Mistral Small 3.2 | 94.9% | Jun 2025 | Lab model card | |
| 14 | Llama 4 Maverick | 94.4% | Apr 2025 | Lab model card | |
| 15 | Nemotron Nano 12B v2 VL | 94.4% | Oct 2025 | Lab model card | |
| 16 | Claude 3.5 Sonnet (October 2024) | 94.2% | Oct 2024 | Lab model card | |
| 17 | Cosmos3-Nano | 94.2% | May 2026 | Lab model card | |
| 18 | Mistral Small 3.1 | 94.1% | Mar 2025 | Lab model card | |
| 19 | Grok-2 Aug 2024 | 93.6% | Aug 2024 | Lab model card | |
| 20 | ERNIE-4.5-VL-28B-A3B-Thinking | 93.6% | Nov 2025 | Lab model card | |
| 21 | Phi-4-multimodal | 93.2% | Feb 2025 | Lab model card | |
| 22 | Gemini 1.5 Pro May 2024 | 93.1% | May 2024 | Lab model card | |
| 23 | GPT-4o May 2024 | 92.8% | May 2024 | Lab model card | |
| 24 | North Micro Vision Instruct | 92.1% | Aug 2026 | Lab model card | |
| 25 | LongCat-Flash-Omni | 91.8% | Oct 2025 | Lab model card | |
| 26 | Llama-3.1-Nemotron-Nano-VL-8B-V1 | 91.2% | Jun 2025 | Lab model card | |
| 27 | LFM2.5-VL-3B | 91.1% | Aug 2026 | Lab model card | |
| 28 | Gemini 1.0 Ultra | 90.9% | Dec 2023 | Lab model card | |
| 29 | Claude 3 Sonnet | 89.5% | Mar 2024 | Lab model card | |
| 30 | Claude 3 Opus | 89.3% | Mar 2024 | Lab model card | |
| 31 | Claude 3 Haiku | 88.8% | Mar 2024 | Lab model card | |
| 32 | Gemma 3 12B | 87.1% | Mar 2025 | Lab model card | |
| 33 | Cosmos3-Edge | 86.8% | May 2026 | Lab model card | |
| 34 | Gemma 3 27B | 86.6% | Mar 2025 | Lab model card | |
| 35 | Gemini 1.5 Pro Feb 2024 | 86.5% | Feb 2024 | Lab model card | |
| 36 | Gemma 3 4B | 75.8% | Mar 2025 | Lab model card |
6 models have held number one in labs’ own numbers since 2023
Lab claims count from the day the card came out, independent runs from the model’s release.
Google DeepMind OpenAI Anthropic Qwen ByteDance Seed
Source: Labs’ model cards
Google DeepMind’s own number for Gemini 1.0 Ultra passed 85% in December 2023
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Reported by the lab in its model card
Best so far
Source: Labs’ model cards
3 labs put it in a launch post in the last six months
Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| Sep 10, 2026 | DeepSeek V4.1 Flash | 95.6 | DocVQA (LLM-Judge) |
| Aug 12, 2026 | LFM2.5-VL-3B | 91.1 | DocVQA (val) |
| Aug 12, 2026 | North Micro Vision Instruct | 92.1 | DocVQA VAL |
| Feb 6, 2026 | ERNIE 5.0 | 95.5 | DocVQA (val) |
| Nov 11, 2025 | ERNIE-4.5-VL-28B-A3B-Thinking | 93.6 | DocVQA (val) |
| Oct 31, 2025 | LongCat-Flash-Omni | 91.8 | DocVQA |
| Oct 28, 2025 | Nemotron Nano 12B v2 VL | 94.4 | DocVQA (val) |
| Sep 23, 2025 | Qwen3-VL-235B-A22B | 96.5 | DocVQA_TEST |
| Jul 31, 2025 | Command A Vision | 95.9 | DocVQA |
| Jun 20, 2025 | Mistral Small 3.2 | 94.9 | DocVQA |
| Jun 3, 2025 | Llama-3.1-Nemotron-Nano-VL-8B-V1 | 91.2 | DocVQA val |
| May 12, 2025 | Seed1.5-VL | 96.9 | DocVQA |
| May 7, 2025 | Mistral Medium 3 | 95.3 | DocVQA 0-shot |
| Apr 5, 2025 | Llama 4 Scout | 94.4 | DocVQA (test), anls |
| Apr 5, 2025 | Llama 4 Maverick | 94.4 | DocVQA (test), anls |
| Mar 27, 2025 | Qwen2.5-Omni-7B | 95.2 | DocVQAtest |
| Mar 17, 2025 | Mistral Small 3.1 | 94.1 | DocVQA |
| Mar 16, 2025 | ERNIE 4.5 | DocVQA | |
| Mar 12, 2025 | Gemma 3 27B | 86.6 | DocVQA (IT); PT DocVQA (val) = 85.6 |
| Feb 26, 2025 | Phi-4-multimodal | 93.2 | DocVQA |
| Jan 28, 2025 | Qwen2.5-VL-72B | 96.4 | DocVQA_VAL |
| Oct 22, 2024 | Claude 3.5 Sonnet (October 2024) | 94.2 | DocVQA (test, ANLS score) |
| Aug 13, 2024 | Grok-2 | 93.6 | DocVQA |
| Jun 20, 2024 | Claude 3.5 Sonnet | 95.2 | DocVQA (test, ANLS score) |
| May 14, 2024 | Gemini 1.5 Pro | 93.1 | DocVQA (test), 0-shot (ANLS, with Cloud OCR annotations) |
| May 13, 2024 | GPT-4o | 92.8 | DocVQA (%) (test) |
| Mar 4, 2024 | Claude 3 Opus | 89.3 | DocVQA (test, ANLS score) |
| Feb 15, 2024 | Gemini 1.5 Pro | 86.5 | DocVQA (test), 0-shot (ANLS) |
| Dec 6, 2023 | Gemini 1.0 Ultra | 90.9 | DocVQA (test) |
| Mar 14, 2023 | GPT-4 | 88.4 | DocVQA ANLS score (test) 0-shot (pixel-only) |
23 system cards report it, from 11 labs
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| Sep 10, 2026 | DeepSeek-V4.1-Flash Base model card | 95.6 | DocVQA, LLM judge, 4-shot | Capabilities |
| May 31, 2026 | Cosmos3-Nano technical report | 94.2 | DocVQA, Cosmos3-Nano 8B column, Table 10 | Capabilities |
| May 31, 2026 | Cosmos3-Edge technical report | 86.8 | DocVQA, Cosmos3-Edge 2B column, Table 10 | Capabilities |
| Apr 27, 2026 | Nemotron 3 Nano Omni 30B-A3B technical report | 95.6 | DocVQA (Test), reasoning on, 93.3 reasoning off (Table 7) | Capabilities |
| Feb 4, 2026 | ERNIE 5.0 technical report | 95.5 | DocVQAval (Table 3) | Capabilities |
| Nov 26, 2025 | Qwen3-VL-235B-A22B technical report | 96.5 | DocVQA test (Table 2) | Capabilities |
| Nov 6, 2025 | Nemotron Nano 12B v2 VL technical report | 93.2 | DocVQA (Test), reasoning on (Table 3). Reasoning off 94.7 | Capabilities |
| Oct 31, 2025 | LongCat-Flash-Omni technical report | 91.8 | DocVQA (Table 5) | Capabilities |
| Aug 25, 2025 | InternVL3.5-241B-A28B technical report | 94.9 | DocVQA (test) (Table 4) | Capabilities |
| May 11, 2025 | Seed1.5-VL technical report | 96.9 | DocVQA, thinking (Table 6). Non-thinking 96.7 | Capabilities |
| Apr 5, 2025 | Llama 4 Scout model card | 94.4 | DocVQA (test), 0-shot, anls, instruction tuned (Instruction tuned models table) | Capabilities |
| Apr 5, 2025 | Llama 4 Maverick model card | 94.4 | DocVQA (test), 0-shot, anls, instruction tuned (Instruction tuned models table) | Capabilities |
| Mar 26, 2025 | Qwen2.5-Omni-7B technical report | 95.2 | DocVQA test (Table 5) | Capabilities |
| Mar 25, 2025 | Gemma 3 4B technical report | 75.8 | DocVQA, IT, P&S applied (Table 16) | Capabilities |
| Mar 25, 2025 | Gemma 3 27B technical report | 86.6 | DocVQA, IT, P&S applied (Table 16) | Capabilities |
| Mar 25, 2025 | Gemma 3 12B technical report | 87.1 | DocVQA, IT, P&S applied (Table 16) | Capabilities |
| Mar 3, 2025 | Phi-4-multimodal technical report | 93.2 | DocVQA (test) (Table 1). Speech-prompt ShareGPT4o DocVQA 87.3 (Table 2) | Capabilities |
| Feb 19, 2025 | Qwen2.5-VL-72B technical report | 96.4 | DocVQA test (Table 5) | Capabilities |
| Oct 22, 2024 | Claude 3.5 Sonnet (October 2024) model card | 94.2 | DocVQA (test, ANLS score), 0-shot (Table 5) | Capabilities |
| Jun 20, 2024 | Claude 3.5 Sonnet model card | 95.2 | DocVQA (test, ANLS score), 0-shot (Table 2) | Capabilities |
| Mar 4, 2024 | Claude 3 Sonnet model card | 89.5 | DocVQA (test, ANLS score), 0-shot (Table 3) | Capabilities |
| Mar 4, 2024 | Claude 3 Opus model card | 89.3 | DocVQA (test, ANLS score), 0-shot (Table 3) | Capabilities |
| Mar 4, 2024 | Claude 3 Haiku model card | 88.8 | DocVQA (test, ANLS score), 0-shot (Table 3) | Capabilities |
More in Vision
- MMMU In use 11.5K college-level multimodal questions across 30 subjects. Multiple-choice and open accuracy. GPT-5.1 85.4%
- MMMU-Pro In use Harder MMMU: college-level multimodal questions with more options and vision-only inputs to remove text shortcuts. GPT-5.6 Sol 84.6%
- CharXiv In use Chart understanding on real arXiv figures. Reasoning questions require synthesizing information from complex charts. Claude Fable 5 93.5%
- MATH-Vision (MathVision) In use 3,040 competition math problems with visual context. Answer accuracy. Qwen 3.8 Max 97.7%
- MMStar In use Models answer 1,500 human-curated image questions that cannot be solved without the image, across 6 core capabilities. Scored by accuracy. Qwen 3.8 Max 85.9%
- ChartQA In use Models answer human-written and generated questions about chart images that need visual and arithmetic reasoning. Scored by relaxed accuracy with a 5 percent numeric tolerance. Command A Vision 90.9%