OPEN BENCHMARKS
KI-Modell-Benchmarks
Sprach-, Bild-, Video- und Audiomodelle mit nachvollziehbaren offenen Benchmarks vergleichen.
Kurze, lange und mehrsprachige Audios nutzen verschiedene Testsätze und werden getrennt bewertet; eine niedrigere WER ist besser.
Weitere Daten bei Artificial Analysis ↓Modelle vergleichen
0/32–3 Modelle aus dieser Rangliste wählen. Nur Messwerte mit gleicher Quelle, Version und Testgruppe werden verglichen.
| Rang | Modell | Wert | Anbieter | Quelle |
|---|---|---|---|---|
| 1 | cohere-transcribe-03-2026Open | 5,42 % | CohereLabs | Open ASR Leaderboard |
| 2 | scribe_v1Proprietary | 5,47 % | zoom | Open ASR Leaderboard |
| 3 | granite-4.0-1b-speechOpen | 5,52 % | ibm-granite | Open ASR Leaderboard |
| 4 | canary-qwen-2.5bOpen | 5,63 % | nvidia | Open ASR Leaderboard |
| 5 | granite-speech-3.3-8bOpen | 5,74 % | ibm-granite | Open ASR Leaderboard |
| 6 | Qwen3-ASR-1.7BOpen | 5,76 % | Qwen | Open ASR Leaderboard |
| 7 | scribe_v2Proprietary | 5,83 % | elevenlabs | Open ASR Leaderboard |
| 8 | granite-speech-3.3-2bOpen | 6 % | ibm-granite | Open ASR Leaderboard |
| 9 | Phi-4-multimodal-instructOpen | 6,02 % | microsoft | Open ASR Leaderboard |
| 10 | parakeet-tdt-0.6b-v2Open | 6,05 % | nvidia | Open ASR Leaderboard |
| 11 | universal-3-proProprietary | 6,21 % | assemblyai | Open ASR Leaderboard |
| 12 | avalon-v1-enProprietary | 6,24 % | aquavoice | Open ASR Leaderboard |
| 13 | parakeet-tdt-0.6b-v3Open | 6,32 % | nvidia | Open ASR Leaderboard |
| 14 | canary-1b-flashOpen | 6,35 % | nvidia | Open ASR Leaderboard |
| 15 | stt-2.6b-enOpen | 6,4 % | kyutai | Open ASR Leaderboard |
| 16 | chirp_2Proprietary | 6,42 % | Open ASR Leaderboard | |
| 16 | Qwen3-ASR-0.6BOpen | 6,42 % | Qwen | Open ASR Leaderboard |
| 17 | canary-1bOpen | 6,504 % | nvidia | Open ASR Leaderboard |
| 18 | Voxtral-Small-24B-2507Open | 6,62 % | mistralai | Open ASR Leaderboard |
| 19 | chirp_3Proprietary | 6,63 % | Open ASR Leaderboard | |
| 20 | moonshine-streaming-mediumOpen | 6,66 % | usefulsensors | Open ASR Leaderboard |
| 21 | CrisperWhisperOpen | 6,67 % | nyrahealth | Open ASR Leaderboard |
| 22 | Zipformer-transducer-XL-290MOpen | 6,79 % | soundsgoodai | Open ASR Leaderboard |
| 23 | scribe_v1Proprietary | 6,88 % | elevenlabs | Open ASR Leaderboard |
| 24 | enhancedProprietary | 6,91 % | speechmatics | Open ASR Leaderboard |
| 25 | parakeet-tdt-1.1bOpen | 7,015 % | nvidia | Open ASR Leaderboard |
| 26 | GLM-ASR-Nano-2512Open | 7,03 % | zai-org | Open ASR Leaderboard |
| 27 | Voxtral-Mini-3B-2507Open | 7,05 % | mistralai | Open ASR Leaderboard |
| 28 | canary-180m-flashOpen | 7,12 % | nvidia | Open ASR Leaderboard |
| 28 | fusionProprietary | 7,12 % | revai | Open ASR Leaderboard |
| 29 | parakeet-rnnt-1.1bOpen | 7,124 % | nvidia | Open ASR Leaderboard |
| 30 | canary-1b-v2Open | 7,15 % | nvidia | Open ASR Leaderboard |
| 31 | distil-large-v3.5Open | 7,21 % | distil-whisper | Open ASR Leaderboard |
| 32 | lite-whisper-large-v3-accOpen | 7,23 % | efficient-speech | Open ASR Leaderboard |
| 33 | parakeet-ctc-1.1bOpen | 7,399 % | nvidia | Open ASR Leaderboard |
| 34 | owsm_ctc_v4_1BOpen | 7,42 % | espnet | Open ASR Leaderboard |
| 35 | lite-whisper-large-v3Open | 7,43 % | efficient-speech | Open ASR Leaderboard |
| 36 | whisper-large-v3Open | 7,44 % | openai | Open ASR Leaderboard |
| 37 | parakeet-tdt_ctc-110mOpen | 7,49 % | nvidia | Open ASR Leaderboard |
| 38 | parakeet-rnnt-0.6bOpen | 7,495 % | nvidia | Open ASR Leaderboard |
| 39 | distil-large-v3Open | 7,518 % | distil-whisper | Open ASR Leaderboard |
| 40 | Voxtral-Mini-4B-Realtime-2602Open | 7,68 % | mistralai | Open ASR Leaderboard |
| 41 | parakeet-ctc-0.6bOpen | 7,69 % | nvidia | Open ASR Leaderboard |
| 42 | lite-whisper-large-v3-turbo-accOpen | 7,77 % | efficient-speech | Open ASR Leaderboard |
| 42 | VibeVoice-ASR-HFOpen | 7,77 % | microsoft | Open ASR Leaderboard |
| 43 | whisper-large-v3-turboOpen | 7,83 % | openai | Open ASR Leaderboard |
| 44 | whisper-large-v2Open | 7,835 % | openai | Open ASR Leaderboard |
| 45 | moonshine-streaming-smallOpen | 7,84 % | usefulsensors | Open ASR Leaderboard |
| 46 | distil-large-v2Open | 7,921 % | distil-whisper | Open ASR Leaderboard |
| 47 | whisper-largeOpen | 7,943 % | openai | Open ASR Leaderboard |
| 48 | machineProprietary | 7,95 % | revai | Open ASR Leaderboard |
| 49 | whisper-medium.enOpen | 8,09 % | openai | Open ASR Leaderboard |
| 50 | owsm_ctc_v3.1_1BOpen | 8,12 % | espnet | Open ASR Leaderboard |
| 51 | omniASR-LLM-7B-v2Open | 8,14 % | Open ASR Leaderboard | |
| 52 | lite-whisper-large-v3-fastOpen | 8,16 % | efficient-speech | Open ASR Leaderboard |
| 53 | owsm_ctc_v3.2_ft_1BOpen | 8,25 % | espnet | Open ASR Leaderboard |
| 54 | chirpProprietary | 8,29 % | Open ASR Leaderboard | |
| 55 | stt_en_conformer_ctc_largeOpen | 8,319 % | nvidia | Open ASR Leaderboard |
| 56 | asr-conformer-largescaleasrOpen | 8,48 % | speechbrain | Open ASR Leaderboard |
| 57 | fixie-ai-ultravox-v0_6-llama-3_1-8bOpen | 8,52 % | ultravox | Open ASR Leaderboard |
| 58 | distil-small.enOpen | 8,568 % | distil-whisper | Open ASR Leaderboard |
| 59 | whisper-small.enOpen | 8,59 % | openai | Open ASR Leaderboard |
| 60 | distil-medium.enOpen | 8,765 % | distil-whisper | Open ASR Leaderboard |
| 61 | stt_en_fastconformer_ctc_largeOpen | 8,965 % | nvidia | Open ASR Leaderboard |
| 62 | stt_en_fastconformer_transducer_largeOpen | 9,058 % | nvidia | Open ASR Leaderboard |
| 63 | moonshine-baseOpen | 9,99 % | usefulsensors | Open ASR Leaderboard |
| 64 | stt_en_conformer_transducer_smallOpen | 10,258 % | nvidia | Open ASR Leaderboard |
| 65 | whisper-base.enOpen | 10,318 % | openai | Open ASR Leaderboard |
| 66 | asr-19m-v2-en-32bOpen | 10,61 % | abr-ai | Open ASR Leaderboard |
| 67 | omniASR-CTC-7B-v2Open | 10,65 % | Open ASR Leaderboard | |
| 68 | stt_en_conformer_ctc_smallOpen | 11,159 % | nvidia | Open ASR Leaderboard |
| 69 | moonshine-streaming-tinyOpen | 12 % | usefulsensors | Open ASR Leaderboard |
| 70 | moonshine-tinyOpen | 12,65 % | usefulsensors | Open ASR Leaderboard |
| 71 | whisper-tiny.enOpen | 12,806 % | openai | Open ASR Leaderboard |
| 72 | asr-wav2vec2-librispeechOpen | 14,348 % | speechbrain | Open ASR Leaderboard |
| 73 | wav2vec2-large-960h-lv60-selfOpen | 21,273 % | Open ASR Leaderboard | |
| 74 | mms-1b-allOpen | 22,539 % | Open ASR Leaderboard | |
| 75 | hubert-xlarge-ls960-ftOpen | 22,548 % | Open ASR Leaderboard | |
| 76 | hubert-large-ls960-ftOpen | 22,694 % | Open ASR Leaderboard | |
| 77 | wav2vec2-large-robust-ft-libri-960hOpen | 22,931 % | Open ASR Leaderboard | |
| 78 | data2vec-audio-large-960hOpen | 23,21 % | Open ASR Leaderboard | |
| 79 | wav2vec2-conformer-rope-large-960h-ftOpen | 23,284 % | Open ASR Leaderboard | |
| 80 | wav2vec2-conformer-rel-pos-large-960h-ftOpen | 23,291 % | Open ASR Leaderboard | |
| 81 | wav2vec2-large-960hOpen | 26,773 % | Open ASR Leaderboard | |
| 82 | data2vec-audio-base-960hOpen | 28,303 % | Open ASR Leaderboard | |
| 83 | wav2vec2-base-960hOpen | 29,403 % | Open ASR Leaderboard | |
| 84 | mms-1b-fl102Open | 39,796 % | Open ASR Leaderboard |
Keine passenden gemessenen Modelle.
EXTERNAL REFERENCE
Weitere Daten bei Artificial Analysis
Vergleichen Sie Modelle auch jenseits der Benchmarkwerte. Aktuelle Daten stehen auf den offiziellen Seiten.
Quellen und Methodik
Open ASR LeaderboardAktuell
- Version
- main
- Lizenz
- Apache-2.0
- Aktualisiert
- 2026-10-03T13:00:11.401Z