Open model
NVIDIA Parakeet
by NVIDIA
NVIDIA's Parakeet family of speech recognition models, released with open weights and run through the NeMo toolkit.
- Self-hosted
Key figures
Every figure about NVIDIA Parakeet
Each figure is a row of a research dataset. Open it for the exact words of its source or the exact request it was read from.
On-Device Speech Recognition Benchmark: 25 Languages
- Parakeet TDT v3, Bulgarian2026-10-1913.42 WER %
- Parakeet TDT v3, Czech2026-10-1910.44 WER %
- Parakeet TDT v3, Danish2026-10-1915.78 WER %
- Parakeet TDT v3, German2026-10-194.84 WER %
- Parakeet TDT v3, Greek2026-10-1931.66 WER %
- Parakeet TDT v3, English2026-10-194.48 WER %
- Parakeet TDT v3, Spanish2026-10-192.79 WER %
- Parakeet TDT v3, Estonian2026-10-1914.17 WER %
- Parakeet TDT v3, Finnish2026-10-1910.11 WER %
- Parakeet TDT v3, French2026-10-195.29 WER %
- Parakeet TDT v3, Croatian2026-10-1912.44 WER %
- Parakeet TDT v3, Hungarian2026-10-1914.09 WER %
- Parakeet TDT v3, Italian2026-10-192.52 WER %
- Parakeet TDT v3, Lithuanian2026-10-1917.75 WER %
- Parakeet TDT v3, Latvian2026-10-1917.27 WER %
- Parakeet TDT v3, Maltese2026-10-1922.02 WER %
- Parakeet TDT v3, Dutch2026-10-196.32 WER %
- Parakeet TDT v3, Polish2026-10-198.21 WER %
- Parakeet TDT v3, Portuguese2026-10-193.62 WER %
- Parakeet TDT v3, Romanian2026-10-1910 WER %
- Parakeet TDT v3, Russian2026-10-194.74 WER %
- Parakeet TDT v3, Slovak2026-10-198.45 WER %
- Parakeet TDT v3, Slovenian2026-10-1919.83 WER %
- Parakeet TDT v3, Swedish2026-10-1912.03 WER %
Dictation Punctuation and Capitalization Accuracy
- Parakeet TDT v3, Bulgarian2026-10-1915.69 %
- Parakeet TDT v3, Czech2026-10-1912.75 %
- Parakeet TDT v3, Danish2026-10-1918.54 %
- Parakeet TDT v3, German2026-10-197.06 %
- Parakeet TDT v3, Greek2026-10-1934.83 %
- Parakeet TDT v3, English2026-10-198.18 %
- Parakeet TDT v3, Spanish2026-10-195.44 %
- Parakeet TDT v3, Estonian2026-10-1915.29 %
- Parakeet TDT v3, Finnish2026-10-1912.1 %
- Parakeet TDT v3, French2026-10-198.04 %
- Parakeet TDT v3, Croatian2026-10-1915.52 %
- Parakeet TDT v3, Hungarian2026-10-1916.56 %
- Parakeet TDT v3, Italian2026-10-196.4 %
- Parakeet TDT v3, Lithuanian2026-10-1920.07 %
- Parakeet TDT v3, Latvian2026-10-1918.62 %
- Parakeet TDT v3, Maltese2026-10-1925.96 %
- Parakeet TDT v3, Dutch2026-10-198.42 %
- Parakeet TDT v3, Polish2026-10-1911.07 %
- Parakeet TDT v3, Portuguese2026-10-197.68 %
- Parakeet TDT v3, Romanian2026-10-1912.93 %
- Parakeet TDT v3, Russian2026-10-197.37 %
- Parakeet TDT v3, Slovak2026-10-1910.29 %
- Parakeet TDT v3, Slovenian2026-10-1921.81 %
- Parakeet TDT v3, Swedish2026-10-1914.74 %
Speech Recognition Accuracy: Word Error Rates
- Parakeet TDT 0.6B V2, NVIDIA NVIDIA20266.4 WER %
- nvidia/parakeet-ctc-1.1b20268.22 WER %
- nvidia/parakeet-tdt-0.6b-v220266.43 WER %
- nvidia/parakeet-tdt-0.6b-v220266.44 WER %
- nvidia/parakeet-tdt-0.6b-v320266.51 WER %
- nvidia/parakeet-tdt-0.6b-v320266.38 WER %
- Parakeet TDT v3, Bulgarian (our test)202613.42 WER %
- Parakeet TDT v3, Czech (our test)202610.44 WER %
- Parakeet TDT v3, Danish (our test)202615.78 WER %
- Parakeet TDT v3, German (our test)20264.84 WER %
- Parakeet TDT v3, Greek (our test)202631.66 WER %
- Parakeet TDT v3, English (our test)20264.48 WER %
- Parakeet TDT v3, Spanish (our test)20262.79 WER %
- Parakeet TDT v3, Estonian (our test)202614.17 WER %
- Parakeet TDT v3, Finnish (our test)202610.11 WER %
- Parakeet TDT v3, French (our test)20265.29 WER %
- Parakeet TDT v3, Croatian (our test)202612.44 WER %
- Parakeet TDT v3, Hungarian (our test)202614.09 WER %
- Parakeet TDT v3, Italian (our test)20262.52 WER %
- Parakeet TDT v3, Lithuanian (our test)202617.75 WER %
- Parakeet TDT v3, Latvian (our test)202617.27 WER %
- Parakeet TDT v3, Maltese (our test)202622.02 WER %
- Parakeet TDT v3, Dutch (our test)20266.32 WER %
- Parakeet TDT v3, Polish (our test)20268.21 WER %
Speech Recognition Speed: Real-Time Factor Data
- Parakeet TDT v3, Bulgarian (M2 Pro)202618 RTFx
- Parakeet TDT v3, Czech (M2 Pro)202619 RTFx
- Parakeet TDT v3, Danish (M2 Pro)202621 RTFx
- Parakeet TDT v3, German (M2 Pro)202620 RTFx
- Parakeet TDT v3, Greek (M2 Pro)202619 RTFx
- Parakeet TDT v3, English (M2 Pro)202619 RTFx
- Parakeet TDT v3, Spanish (M2 Pro)202622 RTFx
- Parakeet TDT v3, Estonian (M2 Pro)202622 RTFx
- Parakeet TDT v3, Croatian (M2 Pro)202620 RTFx
- Parakeet TDT v3, Italian (M2 Pro)202621 RTFx
- Parakeet TDT v3, Lithuanian (M2 Pro)202620 RTFx
- Parakeet TDT v3, Maltese (M2 Pro)202621 RTFx
- Parakeet TDT v3, Dutch (M2 Pro)202619 RTFx
- Parakeet TDT v3, Polish (M2 Pro)202619 RTFx
- Parakeet TDT v3, Portuguese (M2 Pro)202622 RTFx
- Parakeet TDT v3, Russian (M2 Pro)202621 RTFx
- Parakeet TDT v3, Slovenian (M2 Pro)202620 RTFx
- Parakeet TDT v3, Swedish (M2 Pro)202618 RTFx
Most Downloaded Speech Recognition Models (Hugging Face)
- nvidia/parakeet-ctc-0.6bread October 19, 202659,612 downloads
- nvidia/parakeet-ctc-1.1bread October 19, 202688,704 downloads
- nvidia/parakeet-rnnt-0.6bread October 19, 202639,187 downloads
- nvidia/parakeet-rnnt-1.1bread October 19, 20266,336 downloads
- nvidia/parakeet-tdt-0.6b-v2read October 19, 2026197,418 downloads
- nvidia/parakeet-tdt-0.6b-v3read October 19, 2026626,738 downloads
- nvidia/parakeet-tdt_ctc-110mread October 19, 20265,645 downloads
- nvidia/parakeet-unified-en-0.6bread October 19, 20264,447 downloads
Other open models and engines
- Breeze ASRMediaTek Research
- CMU Sphinx (PocketSphinx)Carnegie Mellon University
- Cohere TranscribeCohere
- Coqui STTCoqui
- CrisperWhisperNyra Health
- Distil-WhisperHugging Face
- ESPnetESPnet developers
- faster-whisperSYSTRAN
- FluidAudioFluid Inference
- FunASRModelScope
- GigaAMSaluteDevices (Sber)
- GLM-ASRZ.ai
- IBM Granite SpeechIBM
- IndicConformerAI4Bharat
- insanely-fast-whisperVaibhav Srivastav
- ivrit.ai Whisperivrit.ai
- JuliusJulius development team
- k2 and icefallNext-gen Kaldi team
- KaldiKaldi contributors
- KB-WhisperKBLab, National Library of Sweden
- Kotoba-WhisperKotoba Technologies
- Kyutai STTKyutai
- Lightning Whisper MLXMustafa Aljadery
- Meta MMSMeta AI
- MLX WhisperApple MLX team
- MoonshineMoonshine AI
- Mozilla DeepSpeechMozilla
- NB-WhisperNational Library of Norway AI Lab
- NVIDIA CanaryNVIDIA
- NVIDIA NeMoNVIDIA