Open ASR v1.15: Open ASR Leaderboard v1.15 (average WER, 9 sets)
Average word error rate on the Hugging Face Open ASR Leaderboard (hf-audio), data version v1.15.0 (2026-11-06): the equal-weight default average of nine English short-form sets: AMI-Cleaned, Earnings22-Cleaned-AA-chunked, Gigaspeech-Cleaned, SPGISpeech, URGENT24, Voice Arena Monsoon, Voxpopuli-AA-Cleaned, Private (scripted) and Private (conversational). LibriSpeech clean/other left the default average in v1.14.0. Not comparable with the older 11-set wer-open-asr. Lower is better.
Models ranked
1
tracked on this benchmark
Score band
6.1 – 6.1
best → lowest tracked
Snapshot trend
—
need ≥2 snapshots
Leaderboard
Tracked models ranked by Avg WER (%) (lower is better).
Configuration: nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B
Notes: Benchmark-owner board hf-audio/open_asr_leaderboard (Space sha 9e303409, 2026-11-09), English short-form table, data version v1.15.0 (VERSIONS entry '06-11-2026' in init.py; english_short results revision d2aa732f). Row: rank 9, 'nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B', Average WER 6.13, RTFx 472.62; per set AMI-Cleaned 7.78, Earnings22-Cleaned-AA-chunked 5.25, Gigaspeech-Cleaned 7.18, SPGISpeech 1.91, URGENT24 10.94, Voice Arena Monsoon 3.63, Voxpopuli-AA-Cleaned 1.89, Private (scripted) 3.58, Private (conversational) 13.01. Model added in v1.13.1 (2026-10-28). Board v1.15.1 (2026-11-09) is a UI-only versioning/deeplink release; the data is still the v1.15.0 entry. Not comparable with wer-open-asr (older 11-set definition). NVIDIA's card reports 4.71 on 9 public sets (different set mix); board value used. Retrieved 2026-11-09.
How to read this benchmark
This benchmark scores models where lower is better. Use scores for directional filtering and shortlisting, not universal quality ranking; then validate pricing, context window, provider availability, and fit for your workload.
Trust this score when
- There is a fresh timestamped snapshot (or multiple snapshots) for this benchmark.
- The model list covers the same version family you can actually deploy today.
- Top candidates overlap with your required routing and feature requirements.
Be cautious when
- There is only one benchmark snapshot or the dataset appears stale.
- The benchmark metric direction is opposite of your decision objective.
- The score difference between options is narrow and likely within implementation variance.
Related benchmarks
Last reviewed: Nov 9, 2026