accuracy benchmark
Open ASR Leaderboard Overall is an official aggregate Average WER leaderboard published by Hugging Face Audio.
Updated Sep 8, 2026
Lower average wer ranks better on this benchmark.
Rank | Model | Average WER | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelAUARK-ASR-3BAudio8 | Average WER4.76 wer | Percentile100.00% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank02 | ModelOTMOSS-Transcribe-preview-2BOpenmoss Team | Average WER4.87 wer | Percentile98.31% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank03 | ModelOTMOSS-Transcribe-DiarizeOpenmoss Team | Average WER5.17 wer | Percentile96.61% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank04 | ModelCOcohere-transcribe-03-2026Coherelabs | Average WER5.42 wer | Percentile94.92% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank05 | ModelAUARK-ASR-0.6BAudio8 | Average WER5.53 wer | Percentile93.22% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank06 | ModelSOZipformer-cr-ctc-transducer-XL-290MSoundsgoodai | Average WER5.56 wer | Percentile91.53% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank07 | ModelAC | Average WER5.76 wer | Percentile89.83% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank08 | ModelMI | Average WER6.02 wer | Percentile88.14% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank09 | ModelNV | Average WER6.05 wer | Percentile86.44% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank10 | ModelNV | Average WER6.35 wer | Percentile84.75% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank11 | ModelKYstt-2.6b-enKyutai | Average WER6.40 wer | Percentile83.05% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank12 | ModelAC | Average WER6.42 wer | Percentile81.36% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank13 | ModelNV | Average WER6.50 wer | Percentile79.66% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank14 | ModelMAmoonshine-streaming-mediumMoonshine Ai | Average WER6.66 wer | Percentile77.97% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank15 | ModelSOZipformer-transducer-XL-290MSoundsgoodai | Average WER6.71 wer | Percentile76.27% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank16 | ModelNV | Average WER7.02 wer | Percentile74.58% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank17 | ModelZA | Average WER7.03 wer | Percentile72.88% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank18 | ModelAUAudio8-ASR-0.1BAudio8 | Average WER7.03 wer | Percentile71.19% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank19 | ModelMIVoxtral-Mini-3B-2507Mistralai | Average WER7.05 wer | Percentile69.49% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank20 | ModelNV | Average WER7.12 wer | Percentile67.80% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank21 | ModelNV | Average WER7.12 wer | Percentile66.10% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank22 | ModelNV | Average WER7.15 wer | Percentile64.41% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank23 | ModelDWdistil-large-v3.5Distil Whisper | Average WER7.21 wer | Percentile62.71% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank24 | ModelESowsm_ctc_v4_1BEspnet | Average WER7.42 wer | Percentile61.02% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank25 | ModelOP | Average WER7.44 wer | Percentile59.32% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank26 | ModelNV | Average WER7.49 wer | Percentile57.63% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank27 | ModelNV | Average WER7.50 wer | Percentile55.93% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank28 | ModelNV | Average WER7.69 wer | Percentile54.24% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank29 | ModelMI | Average WER7.77 wer | Percentile52.54% | Participants60 | EvidenceB | EvaluatedN/A |
| Rank30 | ModelOP | Average WER7.84 wer | Percentile50.85% | Participants60 | EvidenceB | EvaluatedN/A |
The leading models and scores on this benchmark.
A closer view of the leading scores on this benchmark.
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
What Open ASR Leaderboard Overall measures and how its scores work.
An official aggregate Average WER leaderboard published by Hugging Face Audio.
Average WER.
Scores are shown in wer. This benchmark is verified and has an evidence level of B.
Open ASR Leaderboard. Benchmark scores retain their original unit. Overall score eligibility is shown separately.
Common questions about Open ASR Leaderboard Overall.
ARK-ASR-3B is currently ranked first with 4.76 wer.
The current leaders are ARK-ASR-3B (4.76 wer), MOSS-Transcribe-preview-2B (4.87 wer), and MOSS-Transcribe-Diarize (5.17 wer).
No matched official input price is currently available.
No matched runtime record is currently available.
No. This benchmark measures one defined capability or task. The overall LLMBoard score uses a separate aggregation across eligible benchmark evidence.
Average WER.
No. Lower values rank better for this benchmark.
60 model results are currently shown.
No. This benchmark is shown for reference but does not contribute to the overall score.
Ranking basisThis open asr leaderboard overall AI model leaderboard uses ascending average wer in the benchmark original unit. The leaderboard ranking keeps matched price and speed data separate from benchmark evidence.
Selection summary
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.