New: connect Claude & other AIs to GenAIList over MCP — research the catalog and contribute to the shared knowledge base. Learn how →
DocVQA benchmark
AI model leaderboard for the DocVQA benchmark. Compare how large language models score on DocVQA, see the full ranking, and understand what this AI benchmark measures. Claude 3.5 Sonnet currently leads with 95.200. Visual question answering over scanned business documents.
Leaderboard
| # | Model | Organization | Score | Variant | Source |
|---|---|---|---|---|---|
| #1 | Claude 3.5 Sonnet | Anthropic | 95.200 | cited-nvlm | official ↗ |
| #2 | Mistral Small 3.2 | Mistral AI | 94.860 | instruct-from-hf-readme | official ↗ |
| #3 | Mistral Small 3.1 | Mistral AI | 94.080 | instruct-cited-msmall32 | official ↗ |
| #4 | Pixtral Large | Mistral AI | 93.300 | anls-from-hf-readme | official ↗ |
| #5 | Gemini 1.5 Pro | Google DeepMind | 93.100 | aug-2024-cited-nvlm | official ↗ |
| #6 | GPT-4o (Aug 2024) | OpenAI | 92.800 | cited-nvlm | official ↗ |
| #7 | Aria | Rhymes AI | 92.600 | from-hf-readme | official ↗ |
| #8 | NVLM-D 72B | NVIDIA | 92.600 | megatron-from-hf-readme | official ↗ |
| #9 | Gemini 1.5 Pro | Google DeepMind | 92.300 | anls-cited-pixtral-large | official ↗ |
| #10 | Pixtral 12B | Mistral AI | 90.700 | cited-aria | official ↗ |
| #11 | Llama 3.2 90B | Meta AI | 90.100 | cited-nvlm | official ↗ |
| #12 | Claude 3.5 Sonnet | Anthropic | 88.600 | anls-cited-pixtral-large | official ↗ |
| #13 | GPT-4o (Aug 2024) | OpenAI | 88.500 | anls-cited-pixtral-large | official ↗ |
| #14 | GPT-4V | OpenAI | 88.400 | cited-nvlm | official ↗ |
| #15 | Llama 3.2 90B | Meta AI | 85.700 | cited-pixtral-large | official ↗ |
| #16 | Llama 3.2 11B | Meta AI | 84.400 | cited-aria | official ↗ |
Frequently asked questions about DocVQA
What is the DocVQA benchmark?
Visual question answering over scanned business documents.
How is the DocVQA benchmark scored?
DocVQA is scored using the ANLS metric, where a higher score is better. The maximum achievable score is 1.000. GenAIList aggregates reported scores from model providers and papers into a single ranked leaderboard.
Which AI model scores highest on DocVQA?
As of the latest reported scores on GenAIList, Claude 3.5 Sonnet achieves the highest result on DocVQA with a score of 95.200.
Is a higher DocVQA score better?
Yes. On DocVQA a higher score indicates better performance, so models near the top of the leaderboard are the strongest.