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

math benchmark

DROP Leaderboard

DROP (Discrete Reasoning Over Paragraphs) is a reading comprehension benchmark requiring discrete reasoning over paragraph content. It contains crowdsourced, adversarially-created questions that require resolving references and performing discrete operations like addition, counting, or sorting, demanding comprehensive paragraph understanding beyond paraphrase-and-entity-typing shortcuts.

Updated Aug 17, 2026

Models30
Model coverage30
MetricScore
EvidenceB

On this page

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

DROP Ranking

Higher score ranks better on this benchmark.

30 rows
Columns

Show columns

Sort by
Rank
Model
Score
Percentile
Participants
Evidence
Evaluated
Rank01ModelDEDeepSeek-V3DeepSeekScore91.6%Percentile100.0%Participants30EvidenceCEvaluatedAug 17, 2026
Rank02ModelANClaude 3.5 SonnetAnthropicScore87.1%Percentile96.5%Participants30EvidenceCEvaluatedAug 17, 2026
Rank03ModelANClaude 3.5 SonnetAnthropicScore87.1%Percentile93.1%Participants30EvidenceCEvaluatedAug 17, 2026
Rank04ModelXIMiMo-V2.5-ProXiaomiScore86.3%Percentile89.7%Participants30EvidenceCEvaluatedAug 17, 2026
Rank05ModelOPGPT-4 TurboOpenAIScore86.0%Percentile86.2%Participants30EvidenceCEvaluatedAug 17, 2026
Rank06ModelAMNova ProAmazonScore85.4%Percentile82.8%Participants30EvidenceCEvaluatedAug 17, 2026
Rank07ModelMELlama 3.1 405B InstructMetaScore84.8%Percentile79.3%Participants30EvidenceCEvaluatedAug 17, 2026
Rank08ModelOPGPT-4oOpenAIScore83.4%Percentile75.9%Participants30EvidenceCEvaluatedAug 17, 2026
Rank09ModelANClaude 3.5 HaikuAnthropicScore83.1%Percentile72.4%Participants30EvidenceCEvaluatedAug 17, 2026
Rank10ModelANClaude 3 OpusAnthropicScore83.1%Percentile69.0%Participants30EvidenceCEvaluatedAug 17, 2026
Rank11ModelOPGPT-4OpenAIScore80.9%Percentile65.5%Participants30EvidenceCEvaluatedAug 17, 2026
Rank12ModelAMNova LiteAmazonScore80.2%Percentile62.1%Participants30EvidenceCEvaluatedAug 17, 2026
Rank13ModelOPGPT-4o miniOpenAIScore79.7%Percentile58.6%Participants30EvidenceCEvaluatedAug 17, 2026
Rank14ModelMELlama 3.1 70B InstructMetaScore79.6%Percentile55.2%Participants30EvidenceCEvaluatedAug 17, 2026
Rank15ModelAMNova MicroAmazonScore79.3%Percentile51.7%Participants30EvidenceCEvaluatedAug 17, 2026
Rank16ModelMELongCat-Flash-ChatMeituanScore79.1%Percentile48.3%Participants30EvidenceCEvaluatedAug 17, 2026
Rank17ModelANClaude 3 SonnetAnthropicScore78.9%Percentile44.8%Participants30EvidenceCEvaluatedAug 17, 2026
Rank18ModelANClaude 3 HaikuAnthropicScore78.4%Percentile41.4%Participants30EvidenceCEvaluatedAug 17, 2026
Rank19ModelMIPhi 4MicrosoftScore75.5%Percentile37.9%Participants30EvidenceCEvaluatedAug 17, 2026
Rank20ModelGOGemini 1.5 ProGoogleScore74.9%Percentile34.5%Participants30EvidenceCEvaluatedAug 17, 2026
Rank21ModelOPGPT-3.5 TurboOpenAIScore70.2%Percentile31.0%Participants30EvidenceBEvaluatedAug 17, 2026
Rank22ModelGOGemma 3n E4BGoogleScore60.8%Percentile27.6%Participants30EvidenceCEvaluatedAug 17, 2026
Rank23ModelGOGemma 3n E4B Instructed LiteRT PreviewGoogleScore60.8%Percentile24.1%Participants30EvidenceCEvaluatedAug 17, 2026
Rank24ModelMELlama 3.1 8B InstructMetaScore59.5%Percentile20.7%Participants30EvidenceCEvaluatedAug 17, 2026
Rank25ModelIBGranite 3.3 8B InstructIBMScore59.4%Percentile17.2%Participants30EvidenceCEvaluatedAug 17, 2026
Rank26ModelGOGemma 3n E2BGoogleScore53.9%Percentile13.8%Participants30EvidenceCEvaluatedAug 17, 2026
Rank27ModelGOGemma 3n E2B Instructed LiteRT (Preview)GoogleScore53.9%Percentile10.3%Participants30EvidenceCEvaluatedAug 17, 2026
Rank28ModelIBIBM Granite 4.0 Tiny PreviewIBMScore46.2%Percentile6.9%Participants30EvidenceCEvaluatedAug 17, 2026
Rank29ModelIBGranite 3.3 8B BaseIBMScore36.1%Percentile3.5%Participants30EvidenceCEvaluatedAug 17, 2026
Rank30ModelBAERNIE 4.5BaiduScore28.6%Percentile0.0%Participants30EvidenceCEvaluatedAug 17, 2026

DROP Highlights

The leading models and scores on this benchmark.

Rank #1DeepSeek-V391.6%Rank #2Claude 3.5 Sonnet87.1%Rank #3Claude 3.5 Sonnet87.1%Rank #4MiMo-V2.5-Pro86.3%

DROP Score Distribution

A closer view of the leading scores on this benchmark.

DROP

The Top AI Models for DROP

The first five results on this benchmark, with official price and output speed added where the model identity can be matched.

Ranking basisThis drop 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
    DE
    DeepSeek-V3DeepSeek
    Score
    91.6%
    Speed
    Up to 100 tok/s via DeepSeek

    Strengths

    • Ranks #1 of 30 compared models
    • 100th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures DROP, not total model capability
  2. 02
    AN
    Claude 3.5 SonnetAnthropic
    Score
    87.1%
    Speed
    Up to 101 tok/s via Bedrock

    Strengths

    • Ranks #2 of 30 compared models
    • 97th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures DROP, not total model capability
  3. 03
    AN
    Claude 3.5 SonnetAnthropic
    Score
    87.1%
    Speed
    Up to 101 tok/s via Bedrock

    Strengths

    • Ranks #3 of 30 compared models
    • 93th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures DROP, not total model capability
  4. 04
    XI
    MiMo-V2.5-ProXiaomi
    Score
    86.3%
    Price
    $0.43 input / $0.87 output per 1M tokens
    Speed
    Up to 69 tok/s via Novita

    Strengths

    • Ranks #4 of 30 compared models
    • 90th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures DROP, not total model capability
  5. 05
    OP
    GPT-4 TurboOpenAI
    Score
    86.0%
    Price
    $10 input / $30 output per 1M tokens
    Speed
    Up to 100 tok/s via OpenAI

    Strengths

    • Ranks #5 of 30 compared models
    • 86th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures DROP, not total model capability

Selection summary

Best AI Models for DROP

DeepSeek-V3 currently leads DROP with 91.6%. 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 #1DeepSeek-V391.6% · Up to 100 tok/s via DeepSeekBenchmark rank #2Claude 3.5 Sonnet87.1% · Up to 101 tok/s via BedrockBenchmark rank #3Claude 3.5 Sonnet87.1% · Up to 101 tok/s via Bedrock

What is DROP?

What DROP measures and how its scores work.

DROP (Discrete Reasoning Over Paragraphs) is a reading comprehension benchmark requiring discrete reasoning over paragraph content. It contains crowdsourced, adversarially-created questions that require resolving references and performing discrete operations like addition, counting, or sorting, demanding comprehensive paragraph understanding beyond paraphrase-and-entity-typing shortcuts.

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

Family
DROP
Modality
text
Primary category
math
Score direction
higher
LLMBoard eligible
Yes
Evaluation key
drop|llm-stats-current

Benchmark scores retain their original unit. Overall score eligibility is shown separately.

FAQ

Common questions about DROP.

Which model scores highest on DROP?

DeepSeek-V3 is currently ranked first with 91.6%.

What does DROP measure?

DROP (Discrete Reasoning Over Paragraphs) is a reading comprehension benchmark requiring discrete reasoning over paragraph content. It contains crowdsourced, adversarially-created questions that require resolving references and performing discrete operations like addition, counting, or sorting, demanding comprehensive paragraph understanding beyond paraphrase-and-entity-typing shortcuts.

Is a higher score better?

Yes. Higher values rank better for this benchmark.

How many models are compared?

30 model results are currently shown.

Does this benchmark affect the overall score?

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