HF Model & Benchmark Matrix

24,078 variants · 14,984 scores · 120 benchmarks

parakeet-tdt-0.6b-v3

nvidia/parakeet-tdt-0.6b-v3

kind
original
quant
bf16 (16-bit)
family
misc
modality
automatic-speech-recognition
root model
nvidia/parakeet-tdt-0.6b-v3
provider
nvidia
license
cc-by-4.0
language
en, es, fr, de, bg, hr, cs, da, nl, et, fi, el
context
—
library
transformers
vRAM (GB)
—
size on disk
—
downloads
202,172
likes
1,024
created
—
last modified
2026-06-29T19:00:30+00:00 (2026)
huggingface
https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3
tags
transformersnemosafetensorsparakeet_tdtfeature-extractionautomatic-speech-recognitionspeechaudioTransducerTransformerTDTFastConformerConformerpytorchNeMohf-asr-leaderboardTransformersenesfrdebghrcs

Benchmark scores (10)

benchmarkcategoryscorecontextconfevidence
WER FLEURSspeech-asr11.97low| **Average WER ↓** | *11.97%* | *7.83%* | *11.98%* |
WERspeech-asr11.97**Average WER ↓**medium| **Average WER ↓** | *11.97%* | *7.83%* | *11.98%* |
WER Earnings-22speech-asr11.42low| `parakeet-tdt-0.6b-v3` | 6.34% | 11.31% | 11.42% | 9.59% | 1.93% | 3.59% | 3.97% | 2.75% | 6.14% |
WER AMIspeech-asr11.31low| `parakeet-tdt-0.6b-v3` | 6.34% | 11.31% | 11.42% | 9.59% | 1.93% | 3.59% | 3.97% | 2.75% | 6.14% |
WER GigaSpeechspeech-asr9.59low| `parakeet-tdt-0.6b-v3` | 6.34% | 11.31% | 11.42% | 9.59% | 1.93% | 3.59% | 3.97% | 2.75% | 6.14% |
WER MLSspeech-asr7.83low| **Average WER ↓** | *11.97%* | *7.83%* | *11.98%* |
WERspeech-asr6.34low| `parakeet-tdt-0.6b-v3` | 6.34% | 11.31% | 11.42% | 9.59% | 1.93% | 3.59% | 3.97% | 2.75% | 6.14% |
WER VoxPopulispeech-asr6.14low| `parakeet-tdt-0.6b-v3` | 6.34% | 11.31% | 11.42% | 9.59% | 1.93% | 3.59% | 3.97% | 2.75% | 6.14% |
WER SPGISpeechspeech-asr3.97low| `parakeet-tdt-0.6b-v3` | 6.34% | 11.31% | 11.42% | 9.59% | 1.93% | 3.59% | 3.97% | 2.75% | 6.14% |
WER TED-LIUMspeech-asr2.75low| `parakeet-tdt-0.6b-v3` | 6.34% | 11.31% | 11.42% | 9.59% | 1.93% | 3.59% | 3.97% | 2.75% | 6.14% |

Discussions (0)

none scraped.

Hugging Face model & benchmark data scraped & matrix-ified. 24,078 variants · 14,984 scores · 120 benchmarks.