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 100 of 100.

ModelTaskSizeDownloads (30 days)
lmstudio-community/Qwen3.8-27B-MLX-4bitMLXfrom Qwen/Qwen3.8-27BVision language16.1 GB3.4M
lmstudio-community/Qwen3.8-27B-MLX-8bitMLXfrom Qwen/Qwen3.8-27BVision language29.5 GB3.2M
lmstudio-community/Qwen3.8-27B-MLX-6bitMLXfrom Qwen/Qwen3.8-27BVision language22.8 GB3.2M
lmstudio-community/Qwen3.8-27B-MLX-5bitMLXfrom Qwen/Qwen3.8-27BVision language19.4 GB3.2M
mlx-community/Qwen3.8-27B-4bitMLXfrom Qwen/Qwen3.8-27BVision language16.1 GB789K
lmstudio-community/Qwen3.5-9B-MLX-4bitMLXfrom Qwen/Qwen3.5-9BVision language6 GB698K
lmstudio-community/Qwen3.5-9B-MLX-8bitMLXfrom Qwen/Qwen3.5-9BVision language10.5 GB647K
lmstudio-community/gemma-4-26B-A4B-it-QAT-MLX-4bitMLXVision language15.6 GB354K
lmstudio-community/Qwen3.6-27B-MLX-4bitMLXfrom Qwen/Qwen3.6-27BVision language16.1 GB322K
lmstudio-community/Qwen3.6-27B-MLX-8bitMLXfrom Qwen/Qwen3.6-27BVision language29.5 GB309K
lmstudio-community/Qwen3.6-27B-MLX-5bitMLXfrom Qwen/Qwen3.6-27BVision language19.4 GB308K
lmstudio-community/Qwen3.6-27B-MLX-6bitMLXfrom Qwen/Qwen3.6-27BVision language22.8 GB307K
lmstudio-community/Qwen3.6-35B-A3B-MLX-4bitMLXfrom Qwen/Qwen3.6-35B-A3BVision language20.4 GB279K
lmstudio-community/Qwen3.6-35B-A3B-MLX-8bitMLXfrom Qwen/Qwen3.6-35B-A3BVision language37.7 GB263K
lmstudio-community/Qwen3.6-35B-A3B-MLX-6bitMLXfrom Qwen/Qwen3.6-35B-A3BVision language29.1 GB258K
mlx-community/olmOCR-2-7B-1025-mlx-8bitMLXfrom Qwen/Qwen2.5-VL-7B-InstructVision language9.5 GB190K
lmstudio-community/gemma-4-26B-A4B-it-MLX-4bitMLXfrom google/gemma-4-26B-A4B-itVision language15.6 GB141K
lmstudio-community/gemma-4-26B-A4B-it-MLX-8bitMLXfrom google/gemma-4-26B-A4B-itVision language28 GB136K
lmstudio-community/gemma-4-26B-A4B-it-MLX-6bitMLXfrom google/gemma-4-26B-A4B-itVision language21.8 GB135K
lmstudio-community/gemma-4-26B-A4B-it-MLX-5bitMLXfrom google/gemma-4-26B-A4B-itVision language18.7 GB134K
lmstudio-community/gemma-4-31B-it-MLX-4bitMLXfrom google/gemma-4-31B-itVision language18.4 GB131K
lmstudio-community/gemma-4-31B-it-MLX-8bitMLXfrom google/gemma-4-31B-itVision language33.8 GB131K
lmstudio-community/gemma-4-31B-it-MLX-6bitMLXfrom google/gemma-4-31B-itVision language26.1 GB126K
lmstudio-community/gemma-4-31B-it-MLX-5bitMLXfrom google/gemma-4-31B-itVision language22.3 GB125K
mlx-community/Devstral-Small-2-24B-Instruct-2512-4bitMLXfrom mistralai/Devstral-Small-2-24B-Instruct-2512Vision language15.1 GB69.7K
lmstudio-community/Qwen3-VL-4B-Instruct-MLX-4bitMLXfrom Qwen/Qwen3-VL-4B-InstructVision language3.1 GB64K
lmstudio-community/Qwen3-VL-4B-Instruct-MLX-8bitMLXfrom Qwen/Qwen3-VL-4B-InstructVision language5.1 GB60.9K
lmstudio-community/Qwen3-VL-4B-Instruct-MLX-6bitMLXfrom Qwen/Qwen3-VL-4B-InstructVision language4.1 GB60.7K
lmstudio-community/Qwen3-VL-4B-Instruct-MLX-5bitMLXfrom Qwen/Qwen3-VL-4B-InstructVision language3.6 GB60.5K
lmstudio-community/Qwen3-VL-8B-Instruct-MLX-4bitMLXfrom Qwen/Qwen3-VL-8B-InstructVision language5.8 GB60.4K
lmstudio-community/Qwen3-VL-8B-Instruct-MLX-8bitMLXfrom Qwen/Qwen3-VL-8B-InstructVision language9.9 GB54.8K
lmstudio-community/Qwen3-VL-8B-Instruct-MLX-6bitMLXfrom Qwen/Qwen3-VL-8B-InstructVision language7.8 GB53.8K
lmstudio-community/Qwen3-VL-8B-Instruct-MLX-5bitMLXfrom Qwen/Qwen3-VL-8B-InstructVision language6.8 GB53.7K
lmstudio-community/GLM-4.6V-Flash-MLX-4bitMLXfrom zai-org/GLM-4.6V-FlashVision language7.1 GB48.6K
lmstudio-community/GLM-4.6V-Flash-MLX-8bitMLXfrom zai-org/GLM-4.6V-FlashVision language11.8 GB47.9K
lmstudio-community/GLM-4.6V-Flash-MLX-6bitMLXfrom zai-org/GLM-4.6V-FlashVision language9.4 GB47.4K
mlx-community/Qwen3.6-27B-4bitMLXfrom Qwen/Qwen3.6-27BVision language16.1 GB44.7K
Jundot/Qwen3.8-Flash-Next-oQ4e-mtpMLXfrom Qwen/Qwen3.8-Flash-NextVision language108 GB43.8K
mlx-community/Qwen3.6-35B-A3B-4bitMLXfrom Qwen/Qwen3.6-35B-A3BVision language20.4 GB43.5K
mlx-community/gemma-3-4b-it-qat-4bitMLXfrom OpenGVLab/InternVL3-1B-InstructVision language3 GB40.2K
mlx-community/gemma-4-e4b-it-4bitMLXfrom google/gemma-4-E4B-itVision language5.2 GB34.7K
TensorFold/Qwen3.8-27B-MLX-4bitMLXfrom Qwen/Qwen3.8-27BVision language16.1 GB32.5K
mlx-community/Qwen3.8-27B-8bitMLXfrom Qwen/Qwen3.8-27BVision language29.5 GB28K
mlx-community/Qwen3.5-9B-MLX-4bitMLXfrom Qwen/Qwen3.5-9BVision language6 GB27.9K
Jundot/Qwen3.8-27B-oQ4e-mtpMLXfrom Qwen/Qwen3.8-27BVision language17.5 GB25.4K
mlx-community/gemma-4-e2b-it-4bitMLXfrom google/gemma-4-E2B-itVision language3.6 GB24.7K
mlx-community/Qwen2-VL-2B-Instruct-4bitMLXVision language1.3 GB23.4K
mlx-community/Qwen2.5-VL-7B-Instruct-4bitMLXfrom Qwen/Qwen2.5-VL-7B-InstructVision language5.7 GB23.3K
mlx-community/Qwen3.5-9B-4bitMLXfrom Qwen/Qwen3.5-9B-BaseVision language6 GB22.7K
mlx-community/gemma-3-4b-it-8bitMLXfrom google/gemma-3-4b-ptVision language5.7 GB20.2K
lmstudio-community/gemma-3n-E4B-it-MLX-4bitMLXfrom google/gemma-3n-E4B-itVision language5.9 GB19.8K
lmstudio-community/Mistral-Small-3.2-24B-Instruct-2506-MLX-6bitMLXfrom mistralai/Mistral-Small-3.2-24B-Instruct-2506Vision language20 GB19.5K
lmstudio-community/gemma-3n-E4B-it-MLX-bf16MLXfrom google/gemma-3n-E4B-itVision language15.7 GB19.5K
lmstudio-community/gemma-3n-E4B-it-MLX-8bitMLXfrom google/gemma-3n-E4B-itVision language9.3 GB19.2K
lmstudio-community/gemma-3n-E4B-it-MLX-6bitMLXfrom google/gemma-3n-E4B-itVision language7.6 GB19.1K
mlx-community/gemma-3-27b-it-qat-4bitMLXfrom OpenGVLab/InternVL3-1B-InstructVision language16.9 GB19K
lmstudio-community/Qwen3-VL-30B-A3B-Instruct-MLX-4bitMLXfrom Qwen/Qwen3-VL-30B-A3B-InstructVision language18.3 GB18.2K
mlx-community/gemma-3-12b-it-qat-4bitMLXfrom OpenGVLab/InternVL3-1B-InstructVision language8.1 GB18.1K
mlx-community/Qwen3-VL-4B-Instruct-4bitMLXVision language3.1 GB18.1K
mlx-community/gemma-4-E2B-it-qat-mobileMLXVision language2.4 GB17.8K
lmstudio-community/Qwen3-VL-30B-A3B-Instruct-MLX-8bitMLXfrom Qwen/Qwen3-VL-30B-A3B-InstructVision language33.5 GB17.8K
Jundot/DeepSeek-V4.1-Flash-oQ4e-mtpMLXfrom deepseek-ai/DeepSeek-V4.1-FlashVision language432 GB17.5K
mlx-community/gemma-3-12b-it-4bitMLXfrom google/gemma-3-12b-ptVision language8.1 GB17.5K
Jundot/Qwen3.6-35B-A3B-oQ4e-mtpMLXfrom Qwen/Qwen3.6-35B-A3BVision language22.2 GB17.2K
lmstudio-community/Qwen3-VL-30B-A3B-Instruct-MLX-6bitMLXfrom Qwen/Qwen3-VL-30B-A3B-InstructVision language25.9 GB16.6K
mlx-community/gemma-4-26b-a4b-it-4bitMLXfrom google/gemma-4-26B-A4B-itVision language15.4 GB16.3K
lmstudio-community/Qwen3-VL-30B-A3B-Instruct-MLX-5bitMLXfrom Qwen/Qwen3-VL-30B-A3B-InstructVision language22.1 GB16.3K
Jundot/gemma-4-26B-A4B-it-oQ4e-mtpMLXfrom google/gemma-4-26B-A4B-itVision language17.1 GB15.1K
Jundot/GLM-5.3-Flash-oQ4e-mtpMLXfrom zai-org/GLM-5.3-FlashVision language186 GB15.1K
lmstudio-community/Mistral-Small-3.2-24B-Instruct-2506-MLX-4bitMLXfrom mistralai/Mistral-Small-3.2-24B-Instruct-2506Vision language13.5 GB14.3K
sh0wie/Qwen3.8-Flash-Next-REAP-288-MLX-4bitMLXfrom Qwen/Qwen3.8-Flash-NextVision language73.5 GB14.1K
mlx-community/gemma-4-E4B-it-qat-mobileMLXVision language3.5 GB13.8K
Jundot/gemma-4-31B-it-oQ6e-mtpMLXfrom google/gemma-4-31B-itVision language27.8 GB13.5K
lmstudio-community/Mistral-Small-3.2-24B-Instruct-2506-MLX-8bitMLXfrom mistralai/Mistral-Small-3.2-24B-Instruct-2506Vision language25.9 GB13.3K
mlx-community/llava-v1.6-mistral-7b-4bitMLXVision language4.3 GB12.9K
mlx-community/Qwen3.5-2B-6bitMLXfrom Qwen/Qwen3.5-2B-BaseVision language2.2 GB12.7K
mlx-community/llava-1.5-7b-4bitMLXVision language4 GB12.6K
lmstudio-community/Qwen3.5-4B-MLX-4bitMLXfrom Qwen/Qwen3.5-4BVision language3.1 GB12.5K
prism-ml/Ternary-Bonsai-2-9B-mlx-2bitMLXfrom Qwen/Qwen3.5-9BVision language3.6 GB11.7K
Youssofal/Qwen3.8-Flash-Next-MTPLX-Optimized-SpeedMLXfrom Qwen/Qwen3.8-Flash-NextVision language115 GB10.9K
mlx-community/Qwen3.8-27B-OptiQ-4bitMLXfrom Qwen/Qwen3.8-27BVision language20.7 GB10.5K
mlx-community/Qwen3.5-0.8B-4bitMLXfrom Qwen/Qwen3.5-0.8B-BaseVision language652 MB10.2K
mlx-community/Muse-Glimmer-30B-4bitMLXfrom meta-models/Muse-Glimmer-30BVision language19.4 GB9,843
lmstudio-community/Qwen3.5-4B-MLX-8bitMLXfrom Qwen/Qwen3.5-4BVision language5.2 GB9,180
mlx-community/gemma-4-31b-it-4bitMLXfrom google/gemma-4-31B-itVision language18.4 GB9,075
mlx-community/Qwen3.5-2B-4bitMLXfrom Qwen/Qwen3.5-2B-BaseVision language1.7 GB8,926
prism-ml/Ternary-Bonsai-2-2B-mlx-2bitMLXfrom Qwen/Qwen3.5-2BVision language1.2 GB8,819
mlx-community/Muse-Spark-Open-816B-A42B-4bitMLXfrom meta-models/Muse-Spark-Open-816B-A42BVision language533 GB8,736
mlx-community/gemma-4-12B-it-qat-4bitMLXVision language11 GB8,349
mlx-community/gemma-4-12B-it-4bitMLXVision language6.8 GB8,348
prism-ml/Ternary-Bonsai-2-4B-mlx-2bitMLXfrom Qwen/Qwen3.5-4BVision language2 GB7,483
unsloth/Qwen3.6-35B-A3B-UD-MLX-4bitMLXfrom Qwen/Qwen3.6-35B-A3BVision language21.7 GB7,473
mlx-community/Qwen3.8-Flash-Next-4bitMLXfrom Qwen/Qwen3.8-Flash-NextVision language112 GB7,395
TensorFold/Qwen3.8-Flash-Next-MLX-4bit-MTPMLXfrom Qwen/Qwen3.8-Flash-NextVision language113 GB7,107
mlx-community/Qwen3.5-0.8B-8bitMLXfrom Qwen/Qwen3.5-0.8B-BaseVision language1 GB7,100
orcarouter/GLM-5.3-Flash-MLXMLXfrom zai-org/GLM-5.3-FlashVision language1.1 TB7,088
mlx-community/Qwen3-VL-2B-Instruct-4bitMLXVision language1.8 GB7,056
TensorFold/Qwen3.8-Flash-Next-MLX-2bit-MTPMLXfrom Qwen/Qwen3.8-Flash-NextVision language80.1 GB6,923
mlx-community/DeepSeek-V4.1-Flash-1017-oQ4e-mtpMLXfrom deepseek-ai/DeepSeek-V4.1-Flash-1017Vision language432 GB6,132
mlx-community/Qwen3-VL-8B-Instruct-4bitMLXVision language5.8 GB6,053

The 100 most downloaded are listed here. For the rest, browse them on Hugging Face. See more on Hugging Face

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.