Skip to content

On-device models

Models you can run on iPhone, iPad and Mac, from Hugging Face

Core ML and MLX models published on Hugging Face, plus the models Apple publishes there, sorted by downloads in the last 30 days. Core ML models run on iPhone, iPad and Mac through Apple's Core ML framework; MLX models are made for Apple's MLX library, mostly on Apple silicon Macs (MLX Swift also runs on iPhone and iPad). Each name links to the model's page on Hugging Face.

1,009 Core ML models, 1,693 MLX models and 143 from Apple, updated daily from Hugging Face's public API. Last updated 2026-11-02 (UTC).

New this month

The most liked models created in the last 30 days, across Core ML, MLX and Apple's own.

See all new models

Showing 69 of 69.

ModelTaskSizeDownloads (30 days)
mlx-community/chatterbox-turbo-4bitMLXfrom ResembleAI/chatterbox-turboText to speech417 MB220K
mlx-community/Kokoro-82M-bf16MLXfrom yl4579/StyleTTS2-LJSpeechText to speech389 MB79K
mlx-community/Qwen3-TTS-12Hz-1.7B-Base-bf16MLXText to speech4.5 GB23.3K
mlx-community/Voxtral-4B-TTS-2603-mlx-bf16MLXfrom mistralai/Voxtral-4B-TTS-2603Text to speech8 GB14.2K
mlx-community/Qwen3-TTS-12Hz-0.6B-Base-8bitMLXText to speech2 GB11.5K
mlx-community/Qwen3-TTS-12Hz-1.7B-Base-8bitMLXText to speech3.1 GB11.4K
mlx-community/Qwen3-TTS-12Hz-1.7B-CustomVoice-8bitMLXfrom Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoiceText to speech3.1 GB10.2K
mlx-community/Qwen3-TTS-12Hz-1.7B-VoiceDesign-4bitMLXText to speech2.3 GB9,935
mlx-community/Qwen3-TTS-12Hz-1.7B-VoiceDesign-bf16MLXText to speech4.5 GB9,778
mlx-community/orpheus-3b-0.1-ft-bf16MLXfrom canopylabs/orpheus-3b-0.1-ftText to speech6.6 GB8,191
mlx-community/Qwen3-TTS-12Hz-0.6B-CustomVoice-8bitMLXText to speech2 GB8,061
mlx-community/Qwen3-TTS-12Hz-1.7B-VoiceDesign-8bitMLXText to speech3.1 GB7,794
mlx-community/Qwen3-TTS-12Hz-0.6B-Base-bf16MLXText to speech2.5 GB7,756
mlx-community/VoxCPM2-8bitMLXfrom openbmb/VoxCPM2Text to speech3.2 GB6,836
mlx-community/chatterbox-multilingual-v3MLXfrom ResembleAI/chatterboxText to speech2.7 GB6,764
mlx-community/Qwen3-TTS-12Hz-1.7B-CustomVoice-bf16MLXText to speech4.5 GB6,029
mlx-community/Breeze-TTS-2-mlxMLXfrom BreezeBlue/Breeze-TTS-2Text to speech—5,612
mlx-community/Qwen3-TTS-12Hz-0.6B-CustomVoice-6bitMLXText to speech—5,053
mlx-community/Qwen3-TTS-12Hz-0.6B-Base-4bitMLXText to speech—4,967
mlx-community/Qwen3-TTS-12Hz-1.7B-CustomVoice-4bitMLXText to speech—4,796
mlx-community/Qwen3-TTS-12Hz-1.7B-Base-4bitMLXText to speech—4,697
mlx-community/fish-audio-drama-3-8bitMLXfrom fishaudio/drama-3Text to speech—4,422
mlx-community/VoxCPM2-bf16MLXfrom openbmb/VoxCPM2Text to speech—3,192
mlx-community/fish-audio-s2-pro-8bitMLXfrom fishaudio/s2-proText to speech—3,099
mlx-community/Voxtral-4B-TTS-2603-mlx-4bitMLXfrom mistralai/Voxtral-4B-TTS-2603Text to speech—2,856
mlx-community/chatterbox-fp16MLXfrom ResembleAI/chatterboxText to speech—2,830
mlx-community/chatterbox-turbo-fp16MLXfrom mlx-community/chatterbox-turbo-fp16Text to speech—2,830
mlx-community/Breeze-TTS-2-mlx-4bitMLXfrom BreezeBlue/Breeze-TTS-2Text to speech—2,778
mlx-community/fish-audio-drama-3-4bitMLXfrom fishaudio/drama-3Text to speech—2,628
mlx-community/Qwen3-TTS-12Hz-0.6B-CustomVoice-bf16MLXText to speech—2,535
theoracleguy/Kokoro-82M-bf16MLXfrom yl4579/StyleTTS2-LJSpeechText to speech—2,324
mlx-community/Qwen3-TTS-12Hz-1.7B-CustomVoice-6bitMLXText to speech—2,318
mlx-community/Qwen3-TTS-12Hz-0.6B-CustomVoice-4bitMLXText to speech—2,308
mlx-community/Soprano-80M-bf16MLXText to speech—2,029
mlx-community/csm-1b-8bitMLXText to speech—1,920
mlx-community/Spark-TTS-0.5B-bf16MLXText to speech—1,789
mlx-community/Kokoro-82M-8bitMLXfrom yl4579/StyleTTS2-LJSpeechText to speech—1,777
mlx-community/Voxtral-4B-TTS-2603-mlx-6bitMLXfrom mistralai/Voxtral-4B-TTS-2603Text to speech—1,758
mlx-community/Chatterbox-TTS-4bitMLXfrom ResembleAI/chatterboxText to speech—1,550
mlx-community/fish-audio-s2-pro-bf16MLXfrom fishaudio/s2-proText to speech—1,538
mlx-community/OmniVoice-bfloat16MLXfrom k2-fsa/OmniVoiceText to speech—1,494
mlx-community/VoxCPM2-4bitMLXfrom openbmb/VoxCPM2Text to speech—1,477
mlx-community/Ming-3.0-tts-0.5B-bf16MLXfrom inclusionAI/Ming-3.0-tts-0.5BText to speech—1,423
mlx-community/chatterbox-turbo-8bitMLXfrom ResembleAI/chatterbox-turboText to speech—1,413
mlx-community/Chatterbox-TTS-fp16MLXfrom ResembleAI/chatterboxText to speech—1,337
mlx-community/Breeze-TTS-2-mlx-8bitMLXfrom BreezeBlue/Breeze-TTS-2Text to speech—1,315
mlx-community/orpheus-3b-0.1-ft-4bitMLXfrom canopylabs/orpheus-3b-0.1-ftText to speech—1,156
mlx-community/Soprano-1.1-80M-bf16MLXfrom ekwek/Soprano-1.1-80MText to speech—1,133
vanch007/mlx-indextts2-2.5-8bitMLXfrom IndexTeam/IndexTTS-2.5Text to speech—1,045
mlx-community/Dia-1.6BMLXfrom nari-labs/Dia-1.6BText to speech—1,013
theoracleguy/OmniVoice-bf16MLXText to speech—1,013
mlx-community/Ming-3.0-tts-0.5B-8bitMLXfrom inclusionAI/Ming-3.0-tts-0.5BText to speech—966
mlx-community/OmniVoice-8bitMLXfrom k2-fsa/OmniVoiceText to speech—945
aufklarer/Chatterbox-Multilingual-MLX-fp16MLXfrom ResembleAI/chatterboxText to speech—941
mlx-community/Qwen3-TTS-12Hz-1.7B-VoiceDesign-6bitMLXText to speech—921
mlx-community/chatterbox-4bitMLXfrom ResembleAI/chatterboxText to speech—905
mlx-community/Llama-OuteTTS-1.0-1B-4bitMLXText to speech—895
mlx-community/Qwen3-TTS-12Hz-1.7B-Base-6bitMLXText to speech—860
mlx-community/chatterbox-8bitMLXfrom ResembleAI/chatterboxText to speech—840
ddalcu/Kokoro-82M-MLX-ServeMLXfrom hexgrad/Kokoro-82MText to speech—790
mlx-community/Kokoro-82M-4bitMLXfrom yl4579/StyleTTS2-LJSpeechText to speech—787
theoracleguy/Qwen3-TTS-12Hz-1.7B-VoiceDesign-4bitMLXfrom Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesignText to speech—730
mlx-community/IndexTTS-2.5-fp16MLXfrom IndexTeam/IndexTTS-2.5Text to speech—695
mlx-community/Fun-CosyVoice3-0.5B-2512-fp16MLXfrom FunAudioLLM/Fun-CosyVoice3-0.5B-2512Text to speech—692
appautomaton/fishaudio-s2-pro-8bit-mlxMLXfrom fishaudio/s2-proText to speech—642
theoracleguy/MOSS-TTS-Local-Transformer-v1.5-4bitMLXfrom OpenMOSS-Team/MOSS-TTS-Local-Transformer-v1.5Text to speech—623
mlx-community/csm-1bMLXfrom sesame/csm-1bText to speech—619
theoracleguy/Qwen3-TTS-12Hz-1.7B-Base-bf16MLXText to speech—607
mlx-community/higgs-audio-v2-3B-mlx-q8MLXfrom bosonai/higgs-tts-2-3b-baseText to speech—571

Browse by task

Pages for tasks with at least 5 models, across Core ML, MLX and Apple's own.

More on on-device models

How to run these

  • Core ML: add the .mlpackage to your Xcode project, or download a compiled model when the app first needs it, and load it with the Core ML framework. Some repos ship their own Swift package; the model card says which.
  • MLX: on a Mac with Apple silicon, the mlx-lm and mlx-audio Python packages load these by name; inside an app, MLX Swift runs them on Mac, iPhone and iPad. Mind the memory: in our test a 4-bit 0.5B model peaked at about 0.36 GB on a Mac; larger models need far more.
  • Apple's own: some are Core ML files ready for an app; others are research weights for PyTorch and need converting with coremltools first.

Reading the table

  • Downloads are Hugging Face's count for the last 30 days, not all time. Hugging Face counts requests for a model's main file, so one install can count more than once.
  • Size is the total of the repo's current files on Hugging Face, in the sizes it shows (1 GB = 1,000 MB). Some repos hold several variants of a model, so it can be larger than the one download you need; a dash means it has not been measured yet.
  • Trending is Hugging Face's own score for recent attention. "from" under a name is the model it was made from.
  • Parameters is the parameter count in the model files. A count marked * comes from the model's name (such as 27B) instead: the files gave none, or a much smaller one, as packed low-bit files can.
  • Quantization is read from the model's tags and name, not checked against the files.

Before you use one

The list comes from Hugging Face's public API and is updated daily. We don't test each model. Check the license and read the model card before you ship one in an app: some licenses limit commercial use, and "other" means the terms are on the model card. Models whose names or tags point at adult or safety-removed content are left out, and so are gated models that need a Hugging Face sign-in to download.