README.md

metadata

license: cc-by-4.0
library_name: mlx
tags:
  - mlx
  - nemo
  - parakeet
  - speech-recognition
  - automatic-speech-recognition
  - fp16
  - apple-silicon
  - ios
  - nvidia
  - rnnt
  - tdt
language:
  - bg
  - cs
  - da
  - de
  - el
  - en
  - es
  - et
  - fi
  - fr
  - hr
  - hu
  - it
  - lt
  - lv
  - mt
  - nl
  - pl
  - pt
  - ro
  - ru
  - sk
  - sl
  - sv
  - uk
pipeline_tag: automatic-speech-recognition
base_model: nvidia/parakeet-tdt-0.6b-v2

Parakeet TDT 0.6B v3 - MLX FP16

This is the NVIDIA Parakeet TDT 0.6B model converted to MLX format with FP16 precision, optimized for Apple Silicon inference.

Parakeet TDT is an ASR model based on the Token-and-Duration Transducer (TDT) architecture, trained on large-scale speech data. The v3 variant adds support for 25 European languages beyond the original English-only v2.

Model Details

Property	Value
Base Model	nvidia/parakeet-tdt-0.6b-v2
Parameters	~600M
Format	MLX SafeTensors (FP16)
Model Size	1,196.08 MB
Sample Rate	16,000 Hz
Architecture	FastConformer + TDT
Encoder Hidden	1024
Predictor Hidden	640
Joint Hidden	640
TDT Durations	[0, 1, 2, 3, 4]
Tokenizer	BPE
Language	25 European languages (BG, CS, DA, DE, EL, EN, ES, ET, FI, FR, HR, HU, IT, LT, LV, MT, NL, PL, PT, RO, RU, SK, SL, SV, UK)

Intended Use

This model is optimized for on-device automatic speech recognition (ASR) on Apple Silicon devices (Mac, iPhone, iPad). It supports 25 European languages: BG, CS, DA, DE, EL, EN, ES, ET, FI, FR, HR, HU, IT, LT, LV, MT, NL, PL, PT, RO, RU, SK, SL, SV, UK.

Files

• config.json - Model configuration
• model.safetensors - Model weights in SafeTensors format (FP16)
• tokenizer.model - SentencePiece tokenizer model
• tokenizer.vocab - Tokenizer vocabulary
• vocab.txt - Vocabulary text file

Original Model

• Paper: Parakeet: A Toolkit for Training and Evaluating ASR Models
• Authors: NVIDIA NeMo Team
• License: CC-BY-4.0