Codesota · Benchmark · Open ASR LeaderboardHome/Leaderboards/Audio & Speech/Automatic Speech Recognition/Open ASR Leaderboard
Unknown

Open ASR Leaderboard.

The Hugging Face Open ASR Leaderboard aggregates Word Error Rate and real-time factor across LibriSpeech, AMI, Earnings-22, GigaSpeech, SPGISpeech, TED-LIUM, and VoxPopuli to give a single composite score for English ASR systems. The de-facto modern ASR leaderboard.

Paper Leaderboard
§ 01 · SOTA history

Year over year.

§ 02 · Leaderboard

Results by metric.

Found a wrong score or missing run?
Use row edits to send a sourced correction into moderation.
Add / edit result Report issue

Rtfx

Rtfx is the reported evaluation metric for Open ASR Leaderboard. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.

Higher is better

Trust tiers for Rtfxverifiedpapervendorcommunityunverified

Muted rows were not state of the art when published — an earlier or same-year result already scored better.

RankModelTrustScoreYearLinksFix
01Stt_en_fastconformer_ctc_largeunverified6399.250312023Paper ↗Source ↗Looks wrong?
02Stt_en_conformer_ctc_smallunverified5686.8965032020Paper ↗Code ↗Source ↗Looks wrong?
03Parakeet-tdt_ctc-110munverified5345.142023Paper ↗Code ↗Source ↗Looks wrong?
04NVIDIA NeMo Conformer-CTC Large
RTFx (real-time factor, higher=faster). nvidia/stt_en_conformer_ctc_large. 120M Conformer CTC. Fast inference (RTFx 4295).
verified4295.012023Source ↗Looks wrong?
05Stt_en_conformer_ctc_largeunverified4295.0066532020Paper ↗Code ↗Source ↗Looks wrong?
06Parakeet-ctc-0.6bunverified4281.5298112023Paper ↗Source ↗Looks wrong?
07Stt_en_fastconformer_transducer_largeunverified4097.4323432023Paper ↗Source ↗Looks wrong?
08Parakeet-tdt-0.6b-v2unverified3386.022023Paper ↗Code ↗Source ↗Looks wrong?
09Parakeet TDT 0.6B v2
RTFx (real-time factor, higher=faster). Rank #10. nvidia/parakeet-tdt-0.6b-v2. RTFx 3386 (fastest open model).
verified3386.022024Source ↗Looks wrong?
10Parakeet TDT 0.6B v3
RTFx (real-time factor, higher=faster). Rank #12. nvidia/parakeet-tdt-0.6b-v3.
verified3332.742025Source ↗Looks wrong?
11Parakeet-rnnt-0.6bunverified2815.7245752023Paper ↗Source ↗Looks wrong?
12Parakeet-TDT-1.1B
RTFx (real-time factor, higher=faster). nvidia/parakeet-tdt-1.1b. RTFx 2390.
verified2390.612024Source ↗Looks wrong?
13Canary-180M-Flash
RTFx (real-time factor, higher=faster). nvidia/canary-180m-flash. 182M params, RTFx 1233.
verified1233.582025Source ↗Looks wrong?
14Canary-1B-Flash
RTFx (real-time factor, higher=faster). Rank #13. nvidia/canary-1b-flash. RTFx 1045.
verified1045.752025Source ↗Looks wrong?
15Moonshine-streaming-tinyunverified847.22026Paper ↗Looks wrong?
16Moonshine-tinyunverified753.062024Paper ↗Code ↗Source ↗Looks wrong?
17Canary 1B v2
RTFx (real-time factor, higher=faster). nvidia/canary-1b-v2.
verified7492024Source ↗Looks wrong?
18Wav2vec2-base-960hunverified686.0029072020Paper ↗Code ↗Source ↗Looks wrong?
19Data2vec-audio-base-960hunverified648.1385322022Paper ↗Code ↗Source ↗Looks wrong?
20Wav2vec2-conformer-rope-large-960h-ftunverified607.8694622020Paper ↗Code ↗Source ↗Looks wrong?
21Moonshine-streaming-smallunverified566.332026Paper ↗Source ↗Looks wrong?
22Moonshine-baseunverified565.972024Paper ↗Code ↗Source ↗Looks wrong?
23Cohere Transcribe (Mar 2026)unverified524.882026Paper ↗Looks wrong?
24Wav2vec2-conformer-rel-pos-large-960h-ftunverified522.4568372020Paper ↗Code ↗Source ↗Looks wrong?
25wav2vec 2.0 Large (960h)unverified516.5796592020Paper ↗Code ↗Source ↗Looks wrong?
26Wav2vec2-large-960h-lv60-selfunverified509.3204172020Paper ↗Code ↗Source ↗Looks wrong?
27Wav2vec2-large-robust-ft-libri-960hunverified503.8085612021Paper ↗Code ↗Source ↗Looks wrong?
28Owsm_ctc_v3.1_1Bunverified502.022024Paper ↗Code ↗Source ↗Looks wrong?
29Hubert-large-ls960-ftunverified495.8627042021Paper ↗Code ↗Source ↗Looks wrong?
30Data2vec-audio-large-960hunverified470.1542042022Paper ↗Code ↗Source ↗Looks wrong?
31Asr-wav2vec2-librispeechunverified451.1819762021Paper ↗Code ↗Source ↗Looks wrong?
32Moonshine Streaming Mediumunverified448.152026Paper ↗Source ↗Looks wrong?
33Canary-Qwen-2.5B
RTFx (real-time factor, higher=faster). Rank #4. nvidia/canary-qwen-2.5b. FastConformer encoder + Qwen2 LM.
verified418.282025Source ↗Looks wrong?
34Hubert-xlarge-ls960-ftunverified361.3176542021Paper ↗Code ↗Source ↗Looks wrong?
35Whisper-tiny.enunverified348.1239352022Paper ↗Code ↗Source ↗Looks wrong?
36Distil-small.enunverified331.8934862023Paper ↗Code ↗Source ↗Looks wrong?
37Whisper-base.enunverified320.6738852022Paper ↗Code ↗Source ↗Looks wrong?
38Granite 4.0 1B Speech
RTFx (real-time factor, higher=faster). Rank #3. ibm-granite/granite-4.0-1b-speech. 1B open model.
verified280.022025Source ↗Looks wrong?
39Distil-medium.enunverified279.7331042023Paper ↗Code ↗Source ↗Looks wrong?
40Granite Speech 3.3 2B
RTFx (real-time factor, higher=faster). Rank #8. ibm-granite/granite-speech-3.3-2b.
verified270.572025Source ↗Looks wrong?
41Whisper-small.enunverified268.9148742022Paper ↗Code ↗Source ↗Looks wrong?
42Canary-1B
RTFx (real-time factor, higher=faster). nvidia/canary-1b. FastConformer encoder + T5 decoder.
verified235.342024Source ↗Looks wrong?
43Mms-1b-fl102unverified234.4231742023Paper ↗Code ↗Source ↗Looks wrong?
44Granite Speech 4.1 2Bunverified231.292025Paper ↗Source ↗Looks wrong?
45Mms-1b-allunverified230.7942512023Paper ↗Code ↗Source ↗Looks wrong?
46Distil-large-v3unverified214.4214312023Paper ↗Code ↗Source ↗Looks wrong?
47Distil-Whisper Large v3
RTFx (real-time factor, higher=faster). distil-whisper/distil-large-v3.
verified214.422024Source ↗Looks wrong?
48Distil-large-v2unverified202.9464412023Paper ↗Code ↗Source ↗Looks wrong?
49Distil-Whisper Large v3.5
RTFx (real-time factor, higher=faster). distil-whisper/distil-large-v3.5. 756M params.
verified202.032024Source ↗Looks wrong?
50Whisper Large v3 Turbounverified200.192022Paper ↗Code ↗Source ↗Looks wrong?
51Lite-whisper-large-v3-turbo-accunverified191.712025Paper ↗Code ↗Source ↗Looks wrong?
52Whisper-medium.enunverified182.129162022Paper ↗Code ↗Source ↗Looks wrong?
53Qwen3-ASR-0.6B
RTFx (real-time factor, higher=faster). Rank #15 (tied). Qwen/Qwen3-ASR-0.6B. Smallest Qwen3-ASR variant.
verified166.232025Source ↗Looks wrong?
54Phi-4 Multimodal Instruct
RTFx (real-time factor, higher=faster). Rank #9. microsoft/Phi-4-multimodal-instruct. 6B param open model.
verified151.12025Source ↗Looks wrong?
55Qwen3-ASR-1.7Bunverified147.932026Paper ↗Code ↗Source ↗Looks wrong?
56Whisper Large v3
RTFx (real-time factor, higher=faster). openai/whisper-large-v3. Baseline Whisper entry on the HF Open ASR Leaderboard.
verified145.512023Source ↗Looks wrong?
57Granite Speech 3.3 8B
RTFx (real-time factor, higher=faster). Rank #5. ibm-granite/granite-speech-3.3-8b. 8B open model.
verified145.422025Source ↗Looks wrong?
58GLM-ASR-Nano-2512
RTFx (real-time factor, higher=faster). zai-org/GLM-ASR-Nano-2512. GLM4 2B + audio encoder.
verified145.282024Source ↗Looks wrong?
59Whisper Large v2unverified144.4521022022Paper ↗Code ↗Source ↗Looks wrong?
60Whisper Largeunverified143.7563192022Paper ↗Code ↗Source ↗Looks wrong?
61Lite-whisper-large-v3-fastunverified120.762025Paper ↗Code ↗Source ↗Looks wrong?
62Voxtral Mini 3B
RTFx (real-time factor, higher=faster). mistralai/Voxtral-Mini-3B-2507. 5B effective params.
verified109.862025Source ↗Looks wrong?
63Voxtral-Mini-4B-Realtime-2602unverified93.322026Paper ↗Source ↗Looks wrong?
64Kyutai STT 2.6B (EN)
RTFx (real-time factor, higher=faster). Rank #14. kyutai/stt-2.6b-en. Streaming ASR based on Moshi architecture.
verified88.372024Source ↗Looks wrong?
65CrisperWhisper
RTFx (real-time factor, higher=faster). nyrahealth/CrisperWhisper. Whisper fine-tune with precise word boundaries.
verified84.052024Source ↗Looks wrong?
66SYMPHONY-ASRunverified77.562026Paper ↗Looks wrong?
67Voxtral-Small-24B-2507
RTFx (real-time factor, higher=faster). mistralai/Voxtral-Small-24B-2507. Mistral large multimodal model.
verified54.092025Source ↗Looks wrong?
68VibeVoice-ASR-HFunverified51.82026Paper ↗Source ↗Looks wrong?
69Asr-conformer-loquaciousunverified42.162025Paper ↗Looks wrong?

Wer

Wer is the reported evaluation metric for Open ASR Leaderboard. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.

Lower is better

Trust tiers for Werverifiedpapervendorcommunityunverified

Muted rows were not state of the art when published — an earlier or same-year result already scored better.

RankModelTrustScoreYearLinksFix
01Moonshine-baseunverified9.992024Paper ↗Code ↗Source ↗Looks wrong?
02Whisper-base.enunverified10.31752022Paper ↗Code ↗Source ↗Looks wrong?
03Niagara-19m-batch.enunverified10.472026Paper ↗Looks wrong?
04Stt_en_conformer_ctc_smallunverified11.158752020Paper ↗Code ↗Source ↗Looks wrong?
05Moonshine-streaming-tinyunverified122026Paper ↗Looks wrong?
06Moonshine-tinyunverified12.652024Paper ↗Code ↗Source ↗Looks wrong?
07Whisper-tiny.enunverified12.806252022Paper ↗Code ↗Source ↗Looks wrong?
08Asr-wav2vec2-librispeechunverified14.34752021Paper ↗Code ↗Source ↗Looks wrong?
09Wav2vec2-large-960h-lv60-selfunverified21.27252020Paper ↗Code ↗Source ↗Looks wrong?
10Mms-1b-allunverified22.538752023Paper ↗Code ↗Source ↗Looks wrong?
11Hubert-xlarge-ls960-ftunverified22.54752021Paper ↗Code ↗Source ↗Looks wrong?
12Hubert-large-ls960-ftunverified22.693752021Paper ↗Code ↗Source ↗Looks wrong?
13Wav2vec2-large-robust-ft-libri-960hunverified22.931252021Paper ↗Code ↗Source ↗Looks wrong?
14Data2vec-audio-large-960hunverified23.212022Paper ↗Code ↗Source ↗Looks wrong?
15Wav2vec2-conformer-rope-large-960h-ftunverified23.283752020Paper ↗Code ↗Source ↗Looks wrong?
16Wav2vec2-conformer-rel-pos-large-960h-ftunverified23.291252020Paper ↗Code ↗Source ↗Looks wrong?
17wav2vec 2.0 Large (960h)unverified26.77252020Paper ↗Code ↗Source ↗Looks wrong?
18Data2vec-audio-base-960hunverified28.30252022Paper ↗Code ↗Source ↗Looks wrong?
19Wav2vec2-base-960hunverified29.40252020Paper ↗Code ↗Source ↗Looks wrong?
20Mms-1b-fl102unverified39.796252023Paper ↗Code ↗Source ↗Looks wrong?
§ 04 · Submit a result

Add to the leaderboard.

← Back to Automatic Speech Recognition