stt_en_fastconformer_transducer_large
NVIDIA🇺🇸 USMajor lab·
Audioopen weights
English speech recognition model that outputs lowercase transcripts, built on the FastConformer Transducer architecture with about 114M parameters. The model card reports a WER of 1.8 on LibriSpeech test-clean with greedy decoding.
Summary of the model card
- License
- cc-by-4.0
- Modalities
- text → text
1 variant
| Large | HF |
Sources
HuggingFaceDeutsch
Englisches Spracherkennungsmodell, das Sprache in Kleinbuchstaben transkribiert und auf der FastConformer-Transducer-Architektur mit rund 114M Parametern basiert. Laut Modelcard liegt die WER auf LibriSpeech test-clean bei 1.8 (greedy decoding, Version 1.18.0).
Benchmarks
| Benchmark | Domain | Value | Evidence |
|---|---|---|---|
| Earnings-22Open ASR Leaderboard (Langform) | audio-stt | 25.3 % WER | 68 |
| Open ASR LeaderboardOpen ASR Leaderboard (Langform) | audio-stt | 22.4625 % WER | 58 |
| TED-LIUM 3Open ASR Leaderboard (Langform) | audio-stt | 5.41 % WER | 49 |
Vendor-reported numbers are marked and count with factor 0.7 in rankings. The evidence score says how much a benchmark still tells you today.
Provenance
- Detected
- Aug 25, 2026
- First source
- openasr
- Description
- Summary of the model card
- Editorially reviewed
- —