nvidia / automatic-speech-recognition updated 1 month ago

parakeet-tdt-0.6b-v3

parakeet-tdt-0.6b-v3 is a 600-million-parameter multilingual automatic speech recognition (ASR) model designed for high-throughput speech-to-text transcription. It extends the parakeet-tdt-0.6b-v2 model by expanding language support from English to 25 European languages. The model automatically detects the language of the audio and transcribes it without requiring additional prompting. It is part of a series of model...

Params
630 M
Context
Downloads 30d
577 K
Likes
1,135
Commercial use: allowed cc-by-4.0 Not gated GGUF · SAFETENSORS 25 languages View on Hugging Face ↗

Download history

daily snapshots · 42 days
▲ 111 K in the last 30 days (16.1%)
761 K322 K
Aug 11Aug 25Sep 8Sep 21

Can you run it?

Estimated VRAM at 8K context unless noted. Pick your hardware to see the verdict per quantization.

FileQuantSizeEst. VRAMVerdict on RTX 4090 · 24 GB
parakeet-tdt-0.6b-v3.q8_0.gguf Q8_0 0.7 GB 1.4 GB ✅ Runs comfortably
model.safetensors f32 2.5 GB 3.4 GB ✅ Runs comfortably
Estimate: file size × 1.1 + KV cache at 8K + 0.5 GB overhead. Not a benchmark — how we calculate this.

Specifications

Architecture
ParakeetForTDT
Parameters
630 M
Tensor type
F32
Vocabulary
8,193
Licence
cc-by-4.0
First seen on the Hub
2025-08-04
Training datasets
nvidia/Granary, nemo/asr-set-3.0
FLEURS (reported)
6.79
CoVoST2 (reported)
5.1
GigaSpeech (reported)
9.59
Vox Populi (reported)
6.14
tedlium-v3 (reported)
2.75
Earnings-22 (reported)
11.42
SPGI Speech (reported)
3.97
AMI (Meetings test) (reported)
11.31
LibriSpeech (clean) (reported)
3.59
Multilingual LibriSpeech (reported)
7.5
Added to our catalog
2026-08-11
Compare with any automatic-speech-recognition model