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

accuracy benchmark

Open ASR Leaderboard Overall Leaderboard

Direct model-level leaderboard results from Open ASR Leaderboard.

Updated Aug 18, 2026

Models60
Model coverage60
MetricAverage WER
EvidenceA

On this page

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

Open ASR Leaderboard Overall Ranking

Lower average wer ranks better on this benchmark.

30 of 60 rows
Columns

Show columns

Sort by
Rank
Model
Average WER
Percentile
Participants
Evidence
Evaluated
Rank01ModelAUARK-ASR-3BAudio8Average WER4.76 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank02ModelOTMOSS-Transcribe-preview-2BOpenmoss TeamAverage WER4.87 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank03ModelOTMOSS-Transcribe-DiarizeOpenmoss TeamAverage WER5.17 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank04ModelCOcohere-transcribe-03-2026CoherelabsAverage WER5.42 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank05ModelAUARK-ASR-0.6BAudio8Average WER5.53 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank06ModelSOZipformer-cr-ctc-transducer-XL-290MSoundsgoodaiAverage WER5.56 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank07ModelACQwen3-ASR-1.7BAlibaba Cloud / Qwen TeamAverage WER5.76 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank08ModelMIPhi-4-multimodal-instructMicrosoftAverage WER6.02 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank09ModelNVparakeet-tdt-0.6b-v2NVIDIAAverage WER6.05 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank10ModelNVcanary-1b-flashNVIDIAAverage WER6.35 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank11ModelKYstt-2.6b-enKyutaiAverage WER6.4 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank12ModelACQwen3-ASR-0.6BAlibaba Cloud / Qwen TeamAverage WER6.42 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank13ModelNVcanary-1bNVIDIAAverage WER6.504 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank14ModelMAmoonshine-streaming-mediumMoonshine AiAverage WER6.66 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank15ModelSOZipformer-transducer-XL-290MSoundsgoodaiAverage WER6.71 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank16ModelNVparakeet-tdt-1.1bNVIDIAAverage WER7.015 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank17ModelZAGLM-ASR-Nano-2512Zhipu AIAverage WER7.03 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank18ModelAUAudio8-ASR-0.1BAudio8Average WER7.03 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank19ModelMIVoxtral-Mini-3B-2507MistralaiAverage WER7.05 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank20ModelNVcanary-180m-flashNVIDIAAverage WER7.12 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank21ModelNVparakeet-rnnt-1.1bNVIDIAAverage WER7.124 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank22ModelNVcanary-1b-v2NVIDIAAverage WER7.15 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank23ModelDWdistil-large-v3.5Distil WhisperAverage WER7.21 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank24ModelESowsm_ctc_v4_1BEspnetAverage WER7.42 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank25ModelOPWhisper Large V3OpenAIAverage WER7.44 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank26ModelNVparakeet-tdt_ctc-110mNVIDIAAverage WER7.49 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank27ModelNVparakeet-rnnt-0.6bNVIDIAAverage WER7.495 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank28ModelNVparakeet-ctc-0.6bNVIDIAAverage WER7.69 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank29ModelMIVibeVoice-ASR-HFMicrosoftAverage WER7.77 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank30ModelOPwhisper-large-v2OpenAIAverage WER7.835 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank31ModelMAmoonshine-streaming-smallMoonshine AiAverage WER7.84 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank32ModelOPwhisper-largeOpenAIAverage WER7.943 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank33ModelOPwhisper-medium.enOpenAIAverage WER8.09 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank34ModelESowsm_ctc_v3.1_1BEspnetAverage WER8.12 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank35ModelNVstt_en_conformer_ctc_largeNVIDIAAverage WER8.319 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank36ModelSPasr-conformer-loquaciousSpeechbrainAverage WER8.48 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank37ModelOPwhisper-small.enOpenAIAverage WER8.59 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank38ModelAAniagara-38m-batch.enAbr AiAverage WER8.91 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank39ModelNVstt_en_fastconformer_ctc_largeNVIDIAAverage WER8.965 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank40ModelNVstt_en_fastconformer_transducer_largeNVIDIAAverage WER9.058 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank41ModelMAmoonshine-baseMoonshine AiAverage WER9.99 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank42ModelOPwhisper-base.enOpenAIAverage WER10.318 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank43ModelAAniagara-19m-batch.enAbr AiAverage WER10.47 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank44ModelNVstt_en_conformer_ctc_smallNVIDIAAverage WER11.159 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank45ModelMAmoonshine-streaming-tinyMoonshine AiAverage WER12 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank46ModelMAmoonshine-tinyMoonshine AiAverage WER12.65 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank47ModelOPwhisper-tiny.enOpenAIAverage WER12.806 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank48ModelSPasr-wav2vec2-librispeechSpeechbrainAverage WER14.348 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank49ModelFAwav2vec2-large-960h-lv60-selfFacebookAverage WER21.273 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank50ModelFAmms-1b-allFacebookAverage WER22.539 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank51ModelFAhubert-xlarge-ls960-ftFacebookAverage WER22.548 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank52ModelFAhubert-large-ls960-ftFacebookAverage WER22.694 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank53ModelFAwav2vec2-large-robust-ft-libri-960hFacebookAverage WER22.931 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank54ModelFAdata2vec-audio-large-960hFacebookAverage WER23.21 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank55ModelFAwav2vec2-conformer-rope-large-960h-ftFacebookAverage WER23.284 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank56ModelFAwav2vec2-conformer-rel-pos-large-960h-ftFacebookAverage WER23.291 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank57ModelFAwav2vec2-large-960hFacebookAverage WER26.773 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank58ModelFAdata2vec-audio-base-960hFacebookAverage WER28.303 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank59ModelFAwav2vec2-base-960hFacebookAverage WER29.403 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026
Rank60ModelFAmms-1b-fl102FacebookAverage WER39.796 werPercentileN/AParticipants60EvidenceAEvaluatedAug 18, 2026

Open ASR Leaderboard Overall Highlights

The leading models and scores on this benchmark.

Rank #1ARK-ASR-3B4.76 werRank #2MOSS-Transcribe-preview-2B4.87 werRank #3MOSS-Transcribe-Diarize5.17 werRank #4cohere-transcribe-03-20265.42 wer

Open ASR Leaderboard Overall Score Distribution

A closer view of the leading scores on this benchmark.

Open ASR Leaderboard Overall

The Top AI Models for Open ASR Leaderboard Overall

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

Ranking basisThis open asr leaderboard overall AI model leaderboard uses ascending average wer in the benchmark's original unit. The leaderboard ranking keeps matched price and speed data separate from benchmark evidence.

  1. 01
    AU
    ARK-ASR-3BAudio8
    Average WER
    4.76 wer

    Strengths

    • Ranks #1 of 60 compared models
    • A evidence result

    Considerations

    • This result measures Open ASR Leaderboard Overall, not total model capability
  2. 02
    OT
    MOSS-Transcribe-preview-2BOpenmoss Team
    Average WER
    4.87 wer

    Strengths

    • Ranks #2 of 60 compared models
    • A evidence result

    Considerations

    • This result measures Open ASR Leaderboard Overall, not total model capability
  3. 03
    OT
    MOSS-Transcribe-DiarizeOpenmoss Team
    Average WER
    5.17 wer

    Strengths

    • Ranks #3 of 60 compared models
    • A evidence result

    Considerations

    • This result measures Open ASR Leaderboard Overall, not total model capability
  4. 04
    CO
    cohere-transcribe-03-2026Coherelabs
    Average WER
    5.42 wer

    Strengths

    • Ranks #4 of 60 compared models
    • A evidence result

    Considerations

    • This result measures Open ASR Leaderboard Overall, not total model capability
  5. 05
    AU
    ARK-ASR-0.6BAudio8
    Average WER
    5.53 wer

    Strengths

    • Ranks #5 of 60 compared models
    • A evidence result

    Considerations

    • This result measures Open ASR Leaderboard Overall, not total model capability

Selection summary

Best AI Models for Open ASR Leaderboard Overall

ARK-ASR-3B currently leads Open ASR Leaderboard Overall with 4.76 wer. 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 #1ARK-ASR-3B4.76 werBenchmark rank #2MOSS-Transcribe-preview-2B4.87 werBenchmark rank #3MOSS-Transcribe-Diarize5.17 wer

What is Open ASR Leaderboard Overall?

What Open ASR Leaderboard Overall measures and how its scores work.

Direct model-level leaderboard results from Open ASR Leaderboard.

Scores are shown in wer. This benchmark is verified and has an evidence level of A.

Family
Open ASR Overall
Modality
audio
Primary category
accuracy
Score direction
lower
LLMBoard eligible
Yes
Evaluation key
open-asr-overall|open-asr|current

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

FAQ

Common questions about Open ASR Leaderboard Overall.

Which model scores highest on Open ASR Leaderboard Overall?

ARK-ASR-3B is currently ranked first with 4.76 wer.

What does Open ASR Leaderboard Overall measure?

Direct model-level leaderboard results from Open ASR Leaderboard.

Is a higher average wer better?

No. Lower values rank better for this benchmark.

How many models are compared?

60 model results are currently shown.

Does this benchmark affect the overall score?

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