Parakeet TDT 0.6b v3
NVIDIA · released Aug 4, 2025 · nvidia/parakeet-tdt-0.6b-v3
Speech to textTranscribes a recording into words
- Type
- Open weightsCreative Commons Attribution 4.0
- Languages
- 25
- Size
- 0.6B
Context measured in tokens
Our take
Written Sep 4, 2026Parakeet is a tiny downloadable speech-to-text model from NVIDIA that turns audio into written words. It covers 25 languages and is built for speed rather than peak accuracy, transcribing an hour of audio in under a second on benchmark hardware.
Pick this for batch transcription where speed matters, or when you need a permissively licensed model with broad language coverage. It handles podcasts, video and recorded meetings better than most alternatives, and copes well with accented speech. Skip it if you need hosted inference, the absolute best accuracy on clean read-aloud audio, or speaker separation and timestamps.
The case for it
- Extremely fast: 6,098 times real time on benchmark hardware, an hour of audio in under a second.
- Strong on everyday internet audio: 8% word-error rate on podcasts and video, better than most models.
- Resilient on hard conditions: 9.4% error on recorded meetings and 10.8% on accented speech, both better than most.
- Broad language coverage and permissive licence: 25 languages including Spanish, French and German, with Creative Commons Attribution 4.0 allowing commercial use.
The case against it
- Mediocre on clean read-aloud audio: 1.51% error, worse than most models on that condition.
- Near the middle of the field overall: 5.66% average error across nine test sets.
- No hosted providers currently listed; you must run it yourself.
How good is it?
An open transcription model for turning speech into text, including speakers with a range of accents.
- transcribing speakers with a range of accentsAccented speech · 10th of 76
TranscriptionTurning speech into text3.5 of 5Open ASR WER · 29th of 76
95.1%
Misses roughly one word in 21, averaged over nine English test sets.
6,076×9th of 74
an hour of audio in under a second, on the board's own hardware. Your machine will differ.
25
Listed on the model card. The accuracy above is English only.
Percentage of words wrong on each set, lower better. Bars are scaled to this model's own worst case; the placing beneath each rate is against every model measured on that set.
Which languages ↓ ↑
English · Spanish · French · German · Bulgarian · Croatian · Czech · Danish · Dutch · Estonian · Finnish · Greek · Hungarian · Italian · Latvian · Lithuanian · Maltese · Polish · Portuguese · Romanian · Slovak · Slovenian · Swedish · Russian · Ukrainian
The figures above come from the Open ASR Leaderboard, an independent public test that runs every model on the same recordings. It is the only measurement of transcription quality we know of, so there are no other scores to show.
Each of these is the same transcription job on a different kind of recording, so together they say where it holds up and where it slips — not how closely it follows an instruction.
Every published score for this model9 scoresEvery figure we hold, from 9 boards, with who ran it and a link to the source — including the boards no rating above is built on.
Can you run it yourself?
Comfortable fit
GeForce RTX 4090 · 24 GB
Room to spare. 21.2 GB spare means a 10% error in the size would not change the answer.
Apple M1 Pro (16-core GPU) · 32 GB
Room to spare. 22.4 GB spare means a 10% error in the size would not change the answer.
Apple M1 (8-core GPU, 8GB unified) · 8 GB
Room to spare. 4.4 GB spare means a 10% error in the size would not change the answer.
These cards answer whether Parakeet TDT 0.6b v3 loads, not how fast it transcribes. Throughput figures for a transcription model come from its text decoder, so treat this as a fit answer rather than a speed one.
Memory use by level
Against a 24 GB card.
What is quantisation? →This model on every device we track71 devicesThe Q4 build most people download, on each device: what the weights come to, how much context the memory leaves, and whether it runs. Smallest device that runs it first. This is a fit answer: whether it loads, not how fast it transcribes.
Models people weigh against Parakeet TDT 0.6b v3
When we formed this view
Recent changes
What moved
first indexed by our pipelineEach date is the day we first saw the change, or the day the maker announced it.
What we do not know about this model yet
- We hold no measured file for it, so all 3 sizes on this page are calculated from the parameter count.
- Nothing we hold says whether an endpoint streams, so we do not show it either way.
Licence and identifiers
What the licence allowsCreative Commons Attribution 4.0, what it allows commercially, and the identifiers you need to pull this model — its Hugging Face repo, our slug and a machine-readable card.
Licence
Creative Commons Attribution 4.0
Permissive content license: any use with attribution. Common for datasets and some model weights.
Identifiers
- Hugging Face
- nvidia/parakeet-tdt-0.6b-v3
- Architecture
- Dense
- Takes in, gives back
- Audio in, text out
- Catalogue slug
- nvidia-parakeet-tdt-0-6b-v3