Downloads · 30 days
48
33% of all-time downloads
aitytech/Parakeet-TDT-0.6B-v3-MLX
Parakeet-TDT-0.6B-v3-MLX is a automatic speech recognition model from aitytech. Use it when you need speech turned into text. It is set up for mlx. The card lists the license as cc-by-4.0.
MLX-format weights for nvidia/parakeet-tdt-0.6b-v3.
Downloads · 30 days
48
33% of all-time downloads
All-time downloads
147
Public
Parameters
627M
2.5 GB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors2.5 GB · 100%
From the Hugging Face model README
MLX-format weights for nvidia/parakeet-tdt-0.6b-v3.
Cloned from mlx-community/parakeet-tdt-0.6b-v3 for use with qwen3-asr-swift ParakeetASR module.
| Property | Value |
|---|---|
| Parameters | 600M |
| Format | MLX safetensors (FP32) |
| Size | ~2.3 GB |
| Architecture | FastConformer + Token-and-Duration Transducer (TDT) |
| Languages | 25 (EN, BG, HR, CS, DA, NL, ET, FI, FR, DE, EL, HU, IT, LV, LT, MT, PL, PT, RO, SK, SL, ES, SV, RU, UK) |
| Timestamps | Native word-level via TDT decoder |
With qwen3-asr-swift:
import ParakeetASR
let model = try await ParakeetASRModel.fromPretrained(
modelId: "aitytech/Parakeet-TDT-0.6B-v3-MLX"
)
let text = try model.transcribeAudio(audioSamples, sampleRate: 16000)
CC-BY-4.0 (inherited from NVIDIA)