Open ASR Leaderboard leaderboard: who is #1

Name: Open ASR Leaderboard rankings
Creator: Hugging Face
License: https://creativecommons.org/licenses/by/4.0/

Hugging Face · Average WER, lower is better · 67 models · published Sep 22, 2026

Hugging Face

Top 25 of 67 models, best variant per model as published

#	Model	Score
1	scribe v2 pro (zoom)as “zoom/scribe_v2_pro”	3.59 WER
2	azure-speech-07-2026 (microsoft)as “microsoft/azure-speech-07-2026”	3.81 WER
3	multilingual (modulate)as “modulate/multilingual”	3.84 WER
4	scribe v2 (elevenlabs)as “elevenlabs/scribe_v2”	3.97 WER
5	resonant-1 (reson8)as “reson8/resonant-1”	4.01 WER
6	resonant-1-flash (reson8)as “reson8/resonant-1-flash”	4.02 WER
7	scribe v1 (zoom)as “zoom/scribe_v1”	4.15 WER
8	azure-speech-06-2026 (microsoft)as “microsoft/azure-speech-06-2026”	4.26 WER
8	asr-k1 (preview) (sophea)as “sophea/asr-k1 (preview)”	4.26 WER
10	Qwen3-ASR-1.7B-hf (Qwen)as “Qwen/Qwen3-ASR-1.7B-hf”	4.31 WER
11	Hojo-ASR-V1 (HojoAI)as “HojoAI/Hojo-ASR-V1”	4.33 WER
12	universal-3-5-pro (assemblyai)as “assemblyai/universal-3-5-pro”	4.34 WER
13	higgs-audio-v3-stt (bosonai)as “bosonai/higgs-audio-v3-stt”	4.39 WER
14	pulse (smallestai)as “smallestai/pulse”	4.41 WER
15	canary-qwen-2.5b (nvidia)as “nvidia/canary-qwen-2.5b”	4.43 WER
16	ARK-ASR-3B (AutoArk-AI)as “AutoArk-AI/ARK-ASR-3B”	4.47 WER
17	ARK-ASR-0.6B (AutoArk-AI)as “AutoArk-AI/ARK-ASR-0.6B”	4.56 WER
18	solaria-3 (gladia)as “gladia/solaria-3”	4.57 WER
19	granite-speech-4.1-2b (ibm-granite)as “ibm-granite/granite-speech-4.1-2b”	4.62 WER
20	MOSS-Transcribe-Diarize (OpenMOSS-Team)as “OpenMOSS-Team/MOSS-Transcribe-Diarize”	4.64 WER
21	cohere-transcribe-03-2026 (CohereLabs)as “CohereLabs/cohere-transcribe-03-2026”	4.67 WER
22	granite-speech-4.1-2b-nar (ibm-granite)as “ibm-granite/granite-speech-4.1-2b-nar”	4.68 WER
23	parakeet-tdt-0.6b-v2 (nvidia)as “nvidia/parakeet-tdt-0.6b-v2”	4.70 WER
24	granite-speech-5.0-470m-turboctc-nc (ibm-granite)as “ibm-granite/granite-speech-5.0-470m-turboctc-nc”	4.78 WER
25	muse-voice-transcribe (meta)as “meta/muse-voice-transcribe”	4.84 WER