llmboard.aiAI model intelligence
Home

Model Rankings

OverallOpen ModelsAgentCodingReasoningMathKnowledgeInstruction FollowingTextVision
Image GenerationImage Editing
Video GenerationImage to VideoVideo Editing
Text to SpeechSpeech to Text
Embeddings

Efficiency

Chat Token PricingImage PricingVideo PricingAudio Pricing
Chat Speed & LatencyProvider Reliability

Benchmarks

GPQAMMLU-ProAIME 2025SWE-Bench VerifiedMMLUHumanity's Last ExamLiveCodeBenchMATHHumanEvalMMMU-Pro
All Benchmarks

Tools

Model DirectoryCompare Models

Scoring & Data

Scoring & Data
393 models668 benchmarks

Leaderboard Center

Overall RankingCodingCore BenchmarksPrice & ValueRuntime Performance

Modalities

All ModelsImage GenerationImage EditingVideo GenerationImage-to-VideoVideo EditingText-to-SpeechSpeech-to-TextEmbeddings

Data & Methods

Scoring MethodAll BenchmarksReasoningMath

Vendors

All VendorsOpenAIAnthropicGoogle
llmboard.aiCopyright 2026 llmboard.ai

multimodal benchmark

DocVQA Leaderboard

A dataset for Visual Question Answering on document images containing 50,000 questions defined on 12,000+ document images. The benchmark tests AI's ability to understand document structure and content, requiring models to comprehend document layout and perform information retrieval to answer questions about document images.

Updated Aug 17, 2026

Models28
Model coverage28
MetricScore
EvidenceB

On this page

  • Ranking
  • Highlights
  • Distribution
  • Top models
  • About
  • FAQ

DocVQA Ranking

Higher score ranks better on this benchmark.

28 rows
Columns

Show columns

Sort by
Rank
Model
Score
Percentile
Participants
Evidence
Evaluated
Rank01ModelACQwen2.5 VL 72B InstructAlibaba Cloud / Qwen TeamScore96.4%Percentile100.0%Participants28EvidenceCEvaluatedAug 17, 2026
Rank02ModelACQwen2.5 VL 7B InstructAlibaba Cloud / Qwen TeamScore95.7%Percentile96.3%Participants28EvidenceCEvaluatedAug 17, 2026
Rank03ModelANClaude 3.5 SonnetAnthropicScore95.2%Percentile92.6%Participants28EvidenceCEvaluatedAug 17, 2026
Rank04ModelACQwen2.5-Omni-7BAlibaba Cloud / Qwen TeamScore95.2%Percentile88.9%Participants28EvidenceCEvaluatedAug 17, 2026
Rank05ModelMAMistral Small 3.2 24B InstructMistral AIScore94.9%Percentile85.2%Participants28EvidenceCEvaluatedAug 17, 2026
Rank06ModelACQwen2.5 VL 32B InstructAlibaba Cloud / Qwen TeamScore94.8%Percentile81.5%Participants28EvidenceCEvaluatedAug 17, 2026
Rank07ModelMELlama 4 MaverickMetaScore94.4%Percentile77.8%Participants28EvidenceCEvaluatedAug 17, 2026
Rank08ModelMELlama 4 ScoutMetaScore94.4%Percentile74.1%Participants28EvidenceCEvaluatedAug 17, 2026
Rank09ModelXAGrok-2xAIScore93.6%Percentile70.4%Participants28EvidenceCEvaluatedAug 17, 2026
Rank10ModelAMNova ProAmazonScore93.5%Percentile66.7%Participants28EvidenceCEvaluatedAug 17, 2026
Rank11ModelDEDeepSeek VL2DeepSeekScore93.3%Percentile63.0%Participants28EvidenceCEvaluatedAug 17, 2026
Rank12ModelMAPixtral LargeMistral AIScore93.3%Percentile59.3%Participants28EvidenceCEvaluatedAug 17, 2026
Rank13ModelXAGrok-2 minixAIScore93.2%Percentile55.6%Participants28EvidenceCEvaluatedAug 17, 2026
Rank14ModelMIPhi-4-multimodal-instructMicrosoftScore93.2%Percentile51.9%Participants28EvidenceCEvaluatedAug 17, 2026
Rank15ModelOPGPT-4oOpenAIScore92.8%Percentile48.1%Participants28EvidenceCEvaluatedAug 17, 2026
Rank16ModelAMNova LiteAmazonScore92.4%Percentile44.4%Participants28EvidenceCEvaluatedAug 17, 2026
Rank17ModelDEDeepSeek VL2 SmallDeepSeekScore92.3%Percentile40.7%Participants28EvidenceCEvaluatedAug 17, 2026
Rank18ModelCONorth Micro Vision InstructCohereScore92.1%Percentile37.0%Participants28EvidenceCEvaluatedAug 17, 2026
Rank19ModelLALFM2.5-VL-3BLiquid AIScore91.1%Percentile33.3%Participants28EvidenceCEvaluatedAug 17, 2026
Rank20ModelMAPixtral-12BMistral AIScore90.7%Percentile29.6%Participants28EvidenceCEvaluatedAug 17, 2026
Rank21ModelMELlama 3.2 90B InstructMetaScore90.1%Percentile25.9%Participants28EvidenceCEvaluatedAug 17, 2026
Rank22ModelDEDeepSeek VL2 TinyDeepSeekScore88.9%Percentile22.2%Participants28EvidenceCEvaluatedAug 17, 2026
Rank23ModelMELlama 3.2 11B InstructMetaScore88.4%Percentile18.5%Participants28EvidenceCEvaluatedAug 17, 2026
Rank24ModelGOGemma 3 12BGoogleScore87.1%Percentile14.8%Participants28EvidenceCEvaluatedAug 17, 2026
Rank25ModelGOGemma 3 27BGoogleScore86.6%Percentile11.1%Participants28EvidenceCEvaluatedAug 17, 2026
Rank26ModelXAGrok-1.5xAIScore85.6%Percentile7.4%Participants28EvidenceCEvaluatedAug 17, 2026
Rank27ModelXAGrok-1.5VxAIScore85.6%Percentile3.7%Participants28EvidenceCEvaluatedAug 17, 2026
Rank28ModelGOGemma 3 4BGoogleScore75.8%Percentile0.0%Participants28EvidenceCEvaluatedAug 17, 2026

DocVQA Highlights

The leading models and scores on this benchmark.

Rank #1Qwen2.5 VL 72B Instruct96.4%Rank #2Qwen2.5 VL 7B Instruct95.7%Rank #3Claude 3.5 Sonnet95.2%Rank #4Qwen2.5-Omni-7B95.2%

DocVQA Score Distribution

A closer view of the leading scores on this benchmark.

DocVQA

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.

Ranking basisThis docvqa AI model leaderboard uses descending score in the benchmark's original unit. The leaderboard ranking keeps matched price and speed data separate from benchmark evidence.

  1. 01
    AC
    Qwen2.5 VL 72B InstructAlibaba Cloud / Qwen Team
    Score
    96.4%

    Strengths

    • Ranks #1 of 28 compared models
    • 100th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures DocVQA, not total model capability
  2. 02
    AC
    Qwen2.5 VL 7B InstructAlibaba Cloud / Qwen Team
    Score
    95.7%
    Price
    $0.35 input / $1.1 output per 1M tokens

    Strengths

    • Ranks #2 of 28 compared models
    • 96th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures DocVQA, not total model capability
  3. 03
    AN
    Claude 3.5 SonnetAnthropic
    Score
    95.2%
    Speed
    Up to 101 tok/s via Bedrock

    Strengths

    • Ranks #3 of 28 compared models
    • 93th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures DocVQA, not total model capability
  4. 04
    AC
    Qwen2.5-Omni-7BAlibaba Cloud / Qwen Team
    Score
    95.2%
    Price
    $0.10 input / $0.40 output per 1M tokens

    Strengths

    • Ranks #4 of 28 compared models
    • 89th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures DocVQA, not total model capability
  5. 05
    MA
    Mistral Small 3.2 24B InstructMistral AI
    Score
    94.9%

    Strengths

    • Ranks #5 of 28 compared models
    • 85th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures DocVQA, not total model capability

Selection summary

Best AI Models for DocVQA

Qwen2.5 VL 72B Instruct currently leads DocVQA with 96.4%. It is the top model on this specific benchmark, while the best LLM for the broader task should also be checked against other benchmarks, price and runtime.

Use this leaderboard with the supporting benchmark results and coverage details above. A leaderboard position summarizes the selected ranking signal; it does not replace workload-specific testing.

Benchmark rank #1Qwen2.5 VL 72B Instruct96.4%Benchmark rank #2Qwen2.5 VL 7B Instruct95.7% · $0.35 input / $1.1 output per 1M tokensBenchmark rank #3Claude 3.5 Sonnet95.2% · Up to 101 tok/s via Bedrock

What is DocVQA?

What DocVQA measures and how its scores work.

A dataset for Visual Question Answering on document images containing 50,000 questions defined on 12,000+ document images. The benchmark tests AI's ability to understand document structure and content, requiring models to comprehend document layout and perform information retrieval to answer questions about document images.

Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.

Family
DocVQA
Modality
multimodal
Primary category
multimodal
Score direction
higher
LLMBoard eligible
Yes
Evaluation key
docvqa|llm-stats-current

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 96.4%.

What does DocVQA measure?

A dataset for Visual Question Answering on document images containing 50,000 questions defined on 12,000+ document images. The benchmark tests AI's ability to understand document structure and content, requiring models to comprehend document layout and perform information retrieval to answer questions about document images.

Is a higher score better?

Yes. Higher values rank better for this benchmark.

How many models are compared?

28 model results are currently shown.

Does this benchmark affect the overall score?

Yes. This benchmark can contribute to the current LLMBoard capability score.