image to text benchmark
DocVQA Leaderboard
DocVQA is a Visual Question Answering benchmark containing 50,000 questions defined on 12,000+ document images.
Updated Sep 5, 2026
Models28
Model coverage28
MetricScore
EvidenceB
DocVQA Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelAC | Score96.40% | Percentile100.00% | Participants28 | EvidenceC | Evaluated |
| Rank02 | ModelAC | Score95.70% | Percentile96.30% | Participants28 | EvidenceC | Evaluated |
| Rank03 | ModelAN | Score95.20% | Percentile92.59% | Participants28 | EvidenceC | Evaluated |
| Rank04 | ModelAC | Score95.20% | Percentile88.89% | Participants28 | EvidenceC | Evaluated |
| Rank05 | ModelMA | Score94.86% | Percentile85.19% | Participants28 | EvidenceC | Evaluated |
| Rank06 | ModelAC | Score94.80% | Percentile81.48% | Participants28 | EvidenceC | Evaluated |
| Rank07 | ModelME | Score94.40% | Percentile77.78% | Participants28 | EvidenceC | Evaluated |
| Rank08 | ModelME | Score94.40% | Percentile74.07% | Participants28 | EvidenceC | Evaluated |
| Rank09 | ModelXA | Score93.60% | Percentile70.37% | Participants28 | EvidenceC | Evaluated |
| Rank10 | ModelAM Nova ProAmazon | Score93.50% | Percentile66.67% | Participants28 | EvidenceC | Evaluated |
| Rank11 | ModelDE | Score93.30% | Percentile62.96% | Participants28 | EvidenceC | Evaluated |
| Rank12 | ModelMA | Score93.30% | Percentile59.26% | Participants28 | EvidenceC | Evaluated |
| Rank13 | ModelXA | Score93.20% | Percentile55.56% | Participants28 | EvidenceC | Evaluated |
| Rank14 | ModelMI | Score93.20% | Percentile51.85% | Participants28 | EvidenceC | Evaluated |
| Rank15 | ModelOP | Score92.80% | Percentile48.15% | Participants28 | EvidenceC | Evaluated |
| Rank16 | ModelAMNova LiteAmazon | Score92.40% | Percentile44.44% | Participants28 | EvidenceC | Evaluated |
| Rank17 | ModelDE | Score92.30% | Percentile40.74% | Participants28 | EvidenceC | Evaluated |
| Rank18 | ModelCO | Score92.10% | Percentile37.04% | Participants28 | EvidenceC | Evaluated |
| Rank19 | ModelLA | Score91.10% | Percentile33.33% | Participants28 | EvidenceC | Evaluated |
| Rank20 | ModelMA | Score90.70% | Percentile29.63% | Participants28 | EvidenceC | Evaluated |
| Rank21 | ModelME | Score90.10% | Percentile25.93% | Participants28 | EvidenceC | Evaluated |
| Rank22 | ModelDE | Score88.90% | Percentile22.22% | Participants28 | EvidenceC | Evaluated |
| Rank23 | ModelME | Score88.40% | Percentile18.52% | Participants28 | EvidenceC | Evaluated |
| Rank24 | ModelGO | Score87.10% | Percentile14.81% | Participants28 | EvidenceC | Evaluated |
| Rank25 | ModelGO | Score86.60% | Percentile11.11% | Participants28 | EvidenceC | Evaluated |
| Rank26 | ModelXA | Score85.60% | Percentile7.41% | Participants28 | EvidenceC | Evaluated |
| Rank27 | ModelXA | Score85.60% | Percentile3.70% | Participants28 | EvidenceC | Evaluated |
| Rank28 | ModelGO | Score75.80% | Percentile0.00% | Participants28 | EvidenceC | Evaluated |
DocVQA Highlights
The leading models and scores on this benchmark.
DocVQA Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for DocVQA
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
What is DocVQA?
What DocVQA measures and how its scores work.
DocVQA is a dataset and benchmark for answering questions about document images.
It measures AI's ability to understand document structure and content, comprehend document layout, and perform information retrieval to answer questions about document images, using Score as a ratio.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- DocVQA
- Modality
- multimodal
- Primary category
- image to text
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- overall
LLM Stats. Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about DocVQA.
Which model scores highest on DocVQA?
Qwen2.5 VL 72B Instruct is currently ranked first with 96.40%.
What are the top three models on DocVQA?
The current leaders are Qwen2.5 VL 72B Instruct (96.40%), Qwen2.5 VL 7B Instruct (95.70%), and Claude 3.5 Sonnet (95.20%).
Which DocVQA model has the lowest official input price?
Nova Lite has the lowest matched official input price at $0.06 input / $0.24 output per 1M tokens.
Which models are fastest among DocVQA results?
The fastest matched records are Llama 4 Scout (776.10 tok/s via Groq), Llama 4 Maverick (307.30 tok/s via Groq), and GPT-4o (132.00 tok/s via OpenAI).
Does the highest score result prove overall model quality?
No. This benchmark measures one defined capability or task. The overall LLMBoard score uses a separate aggregation across eligible benchmark evidence.