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 74 of 74.

ModelTaskSizeDownloads (30 days)
distilbert/distilgpt2Core MLText generation3.2 GB1.5M
tiiuae/falcon-7b-instructCore MLText generation56.6 GB32.1K
Gustavosta/MagicPrompt-Stable-DiffusionCore MLText generation2 GB8,628
enterprise-explorers/Llama-2-7b-chat-coremlCore MLText generation—3,573
vladimir94/dory-1.0-4b-ane-coremlCore MLfrom nvidia/NVIDIA-Nemotron-Labs-Dory1.0-4B-BF16Text generation—1,624
vladimir94/lfm3-omni-2b-ane-coremlCore MLfrom LiquidAI/LFM3-Omni-2BText generation—799
ales27pm/Dolphin3.0-CoreMLCore MLfrom dphn/Dolphin3.0-Llama3.2-3BText generation—686
Norod78/distilgpt2-base-pretrained-heCore MLText generation—658
Writer/palmyra-smallCore MLText generation—638
vstreluk/babelbit-distilgpt2-sn59Core MLText generation—512
Norod78/hebrew-bad_wiki-gpt_neo-tinyCore MLText generation—498
takanori-ishikawa/Qwen3.5-ANE-CoreMLCore MLfrom Qwen/Qwen3.5-0.8BText generation—491
mlboydaisuke/qwen3.5-2B-CoreMLCore MLfrom Qwen/Qwen3.5-2BText generation—444
mlboydaisuke/qwen3.5-0.8B-CoreMLCore MLfrom Qwen/Qwen3.5-0.8BText generation—376
alandotcom/kol-cleanup-360m-coremlCore MLfrom HuggingFaceTB/SmolLM2-360M-InstructText generation—353
getonit/Keyboard-LMCore MLText generation—234
vstreluk/Babelbit-wallet2-sn59-29Core MLText generation—206
okayuji/Qwen3.8-27B-coreml-agentCore MLfrom Qwen/Qwen3.8-27BText generation—205
mlboydaisuke/MiniCPM5-2B-CoreAICore MLfrom openbmb/MiniCPM5-2BText generation—188
aufklarer/Qwen3.5-0.8B-Chat-CoreMLCore MLfrom Qwen/Qwen3.5-0.8BText generation—174
model-attribution-challenge/distilgpt2Core MLText generation—171
mlboydaisuke/MiniCPM5-1B-CoreAICore MLfrom openbmb/MiniCPM5-1BText generation—168
mlboydaisuke/Qwen3.6-35B-A3B-CoreAICore MLfrom Qwen/Qwen3.6-35B-A3BText generation—166
mlboydaisuke/GLM-4.7-Flash-CoreAICore MLfrom zai-org/GLM-4.7-FlashText generation—164
mlboydaisuke/LFM2.5-8B-A1B-CoreAICore MLfrom LiquidAI/LFM2.5-8B-A1BText generation—164
mlboydaisuke/Ornith-1.0-9B-CoreAICore MLfrom ornith-ai/Ornith-1.0-9BText generation—131
FluidInference/short-reply-0.6b-coremlCore MLfrom Qwen/Qwen3-0.6BText generation—107
finnvoorhees/coreml-Qwen2.5-0.5B-Instruct-4bitCore MLText generation—105
runanywhere/LFM2.5-2.6B_ANECore MLfrom LiquidAI/LFM2.5-2.6BText generation—105
FluidInference/qwen2.5-coder-0.5b-coremlCore MLfrom Qwen/Qwen2.5-Coder-0.5B-InstructText generation—94
finnvoorhees/coreml-Llama-3.2-1B-Instruct-4bitCore MLText generation—76
finnvoorhees/coreml-SmolLM2-360M-Instruct-4bitCore MLText generation—68
XDGCC/coreml-Qwen3-0.6BCore MLText generation—55
pywind/coreml-EXAONE-4.0-1.2B-4bitCore MLfrom LGAI-EXAONE/EXAONE-4.0-1.2BText generation—54
finnvoorhees/coreml-Llama-3.2-3B-Instruct-4bitCore MLText generation—53
tianharjuno/Qwen2.5-Coder-1.5B-ANE-ctx2048-LUT6Core MLfrom Qwen/Qwen2.5-Coder-1.5BText generation—52
mlboydaisuke/RWKV7-Goose-1.5B-CoreAICore MLfrom RWKV/RWKV7-Goose-World3-1.5B-HFText generation—49
PiterQ/Bielik-1.5B-v3.0-Instruct-ANE-LUT6-ctx4096Core MLfrom speakleash/Bielik-1.5B-v3.0-InstructText generation—48
tianharjuno/Qwen2.5-Coder-0.5B-ANE-ctx2048-LUT6Core MLfrom Qwen/Qwen2.5-Coder-0.5BText generation—47
NagaYu/halfword-100mCore MLText generation—46
tianharjuno/Qwen2.5-Coder-0.5B-ANE-ctx2048-FP16Core MLfrom Qwen/Qwen2.5-Coder-0.5BText generation—45
mlboydaisuke/FastContext-1.0-4B-CoreAICore MLfrom microsoft/FastContext-1.0-4B-SFTText generation—44
mlboydaisuke/qwen3-1.7b-CoreAI-officialCore MLfrom Qwen/Qwen3-1.7BText generation—44
masahiroid/LFM3-1.7B-CoreMLCore MLfrom LiquidAI/LFM3-1.7BText generation—41
leolaish/Leo-distilGPT-medicalCore MLText generation—40
okayuji/gemma-4-12b-it-coreml-128kCore MLfrom google/gemma-4-12B-itText generation—39
aufklarer/Qwen3-0.6B-Chat-CoreMLCore MLfrom Qwen/Qwen3-0.6BText generation—38
cristianoaredes/orqo-hype-v1Core MLfrom Qwen/Qwen2.5-Coder-7B-InstructText generation—37
finnvoorhees/coreml-Qwen2.5-3B-Instruct-4bitCore MLText generation—37
darkmaniac7/Qwen3-4B-CoreML-ANECore MLfrom Qwen/Qwen3-4BText generation—36
Skyline23/translategemma-4b-it-coremlCore MLfrom google/translategemma-4b-itText generation—35
jiajiahong2134/DLhw2Core MLText generation—35
vasimk/own-gpt-model-privateCore MLText generation—34
anemll/anemll-gemma-3-270m-it-MONO-ctx512-lut6Core MLfrom google/gemma-3-270m-itText generation—30
apple/mistral-coremlCore MLText generation—29
finnvoorhees/coreml-SmolLM2-135M-Instruct-8bitCore MLText generation—29
finnvoorhees/coreml-SmolLM2-135M-Instruct-4bitCore MLText generation—28
fishlessprojects/snowman-1Core MLText generation—27
Reza2kn/MiniCPM5-1B-CoreML-4bitCore MLText generation—26
finnvoorhees/coreml-Qwen2.5-Coder-0.5B-Instruct-4bitCore MLText generation—25
sasn59/Babelbit-hksa02Core MLText generation—25
finnvoorhees/coreml-SmolLM2-1.7B-Instruct-8bitCore MLText generation—23
mlboydaisuke/qwen2.5-0.5b-coremlCore MLfrom Qwen/Qwen2.5-0.5B-InstructText generation—23
nktskr/MamayLM-Gemma-3-4B-IT-ANE-ctx1024Core MLfrom INSAIT-Institute/MamayLM-Gemma-3-4B-IT-v1.0Text generation—23
finnvoorhees/coreml-SmolLM2-1.7B-Instruct-4bitCore MLText generation—22
jonomon/gpt3-kor-small_based_on_gpt2_core_mlCore MLText generation—22
dungnvt/Qwen3-4B-Thinking-2507-LUT444-ctx4096_0.3.4-anemllCore MLfrom Qwen/Qwen3-4B-Thinking-2507Text generation—21
oceanicity/Qwen3-4B-Instruct-CoreML-8bitCore MLfrom oceanicity/Qwen3-4B-Instruct-2507Text generation—21
yacht/Llama-3.2-1B-Instruct-CoreMLCore MLfrom meta-llama/Llama-3.2-1B-InstructText generation—21
JarvanWang/bert-base-uncased-videollama-pretrainedCore MLText generation—20
anziank/grio-qwen2.5-1.5b-instruct-coreml-stateful-int8Core MLfrom Qwen/Qwen2.5-1.5B-InstructText generation—19
finnvoorhees/coreml-SmolLM2-360M-Instruct-8bitCore MLText generation—19
snorkelai/falcon-7b-instruct-pytorch_2.0.0Core MLText generation—18
aufklarer/FunctionGemma-270M-CoreMLCore MLfrom google/functiongemma-270m-itText generation—17

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.