Every AI leaderboard

Leaderboards28Models ranked986

28 leaderboards · 986 models ranked

Open ASR Leaderboard leaderboard: who is #1

Hugging Face · Average WER, lower is better · 62 models · published Nov 6, 2026

Hugging Face

Top 25 of 62 models, best variant per model as published

#ModelScoreCorrect
1New, entered at #1 since Oct 31, 2026scribe v2.1 pro (zoom)as “zoom/scribe_v2.1_pro”4.87 WER
21down from #1 since Oct 31, 2026speech-v3.1 (OtterAI)as “OtterAI/speech-v3.1”5.21 WER
3New, entered at #3 since Nov 7, 2026azure-speech-09-2026 (microsoft)as “microsoft/azure-speech-09-2026”5.55 WER
41down from #3 since Nov 7, 2026audrey-0.7 (p99lab)as “p99lab/audrey-0.7”5.70 WER
51up from #6 since Nov 7, 2026scribe v2 (elevenlabs)as “elevenlabs/scribe_v2”5.93 WER
61down from #5 since Nov 7, 2026multilingual (modulate)as “modulate/multilingual”5.96 WER
71down from #6 since Nov 3, 2026resonant-1 (reson8)as “reson8/resonant-1”5.99 WER
8New, entered at #8 since Nov 7, 2026asr-k1 (sophea)as “sophea/asr-k1”6.11 WER
9New, entered at #9 since Nov 7, 2026NVIDIA-Nemotron-3.5-Transcribe-30B-A4B (nvidia)as “nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B”6.13 WER
101up from #11 since Nov 3, 2026Qwen3-ASR-1.7B-hf (Qwen)as “Qwen/Qwen3-ASR-1.7B-hf”6.15 WER
111down from #10 since Nov 3, 2026universal-3-5-pro (assemblyai)as “assemblyai/universal-3-5-pro”6.21 WER
121up from #13 since Nov 3, 2026pulse (smallestai)as “smallestai/pulse”6.57 WER
131down from #12 since Nov 3, 2026granite-speech-5.0-2b (ibm-granite)as “ibm-granite/granite-speech-5.0-2b”6.68 WER
143up from #17 since Nov 7, 2026cohere-transcribe-03-2026 (CohereLabs)as “CohereLabs/cohere-transcribe-03-2026”6.72 WER
151down from #14 since Nov 7, 2026Hojo-ASR-V1 (HojoAI)as “HojoAI/Hojo-ASR-V1”6.74 WER
151up from #16 since Nov 7, 2026thewhisper-large-v3-turbo (TheStageAI)as “TheStageAI/thewhisper-large-v3-turbo”6.74 WER
172down from #15 since Nov 7, 2026ARK-ASR-0.6B (Edge0)as “Edge0/ARK-ASR-0.6B”6.75 WER
184up from #22 since Nov 3, 2026gpt-transcribe (openai)as “openai/gpt-transcribe”6.80 WER
194down from #15 since Nov 3, 2026symphony (sprag)as “sprag/symphony”6.85 WER
202up from #22 since Nov 7, 2026solaria-3 (gladia)as “gladia/solaria-3”6.86 WER
213up from #24 since Nov 7, 2026parakeet-tdt-0.6b-v2 (fast-gpu-asr) (nvidia)as “nvidia/parakeet-tdt-0.6b-v2 (fast-gpu-asr)”6.91 WER
214up from #25 since Nov 7, 2026MOSS-Transcribe-Diarize (OpenMOSS-Team)as “OpenMOSS-Team/MOSS-Transcribe-Diarize”6.91 WER
235down from #18 since Nov 3, 2026ARK-ASR-3B (Edge0)as “Edge0/ARK-ASR-3B”6.92 WER
232down from #21 since Nov 7, 2026canary-qwen-2.5b (nvidia)as “nvidia/canary-qwen-2.5b”6.92 WER
251up from #26 since Nov 7, 2026parakeet-tdt-0.6b-v2 (nvidia)as “nvidia/parakeet-tdt-0.6b-v2”6.93 WER
About this board and its history

What it measures

Speech recognition word error rate across eight public English test sets. Lower is better.

Rank at each month's end

12345678910SepOctNow1scribe v2.1 pro (zoom)2speech-v3.1 (OtterAI)3azure-speech-09-2026 (mi4audrey-0.7 (p99lab)5scribe v2 (elevenlabs)6multilingual (modulate)14
scribe v2.1 pro (zoom)speech-v3.1 (OtterAI)azure-speech-09-2026 (microsoft)audrey-0.7 (p99lab)scribe v2 (elevenlabs)multilingual (modulate)zoom-scribe-v2-promicrosoft-azure-speech-07-2026

Other speech boards

Source

Scores as published by Hugging Face on the capture date (2026-11-09). Source: hf-audio results CSV on Hugging Face.

huggingface.co/spaces/hf-audio/open_asr_leaderboard