Skip to content

All on-device models

Text classification models for iPhone and Mac (Core ML & MLX)

50 models for this task are tagged Core ML or MLX on Hugging Face, or come from Apple: 45 Core ML, 5 MLX and 0 other Apple models. The table lists the most downloaded in the last 30 days.

Last updated 2026-11-02 (UTC).

Showing 50 of 50.

ModelSizeDownloads (30 days)
mlx-community/Qwen3.5-0.8B-decisionMLXfrom Qwen/Qwen3.5-0.8B—1,857
mlx-community/gemma-4-e2b-it-OptiQ-4bit-decisionMLXfrom mlx-community/gemma-4-e2b-it-OptiQ-4bit—1,037
mlx-community/pplx-decider-v1.1-27b-4bitMLXfrom perplexity-ai/pplx-decider-v1.1-27b—1,011
cavi-ai/laya-MLX-8bitMLXfrom convaiinnovations/laya—736
chaoliangUNSW/Jev-Style-2B-Decision-v3-MLXMLXfrom chaoliangUNSW/Jev-Style-2B-Decision-v3—658
FluidInference/gliner2-5-decide-coremlCore MLfrom fastino/GLiNER2.5-Decide—312
JohnP1/d1a-e4b-coremlCore MLfrom google/gemma-4-E4B—259
FluidInference/clef-text-0.6b-coremlCore MLfrom FluidInference/clef-text-0.6b—238
FluidInference/clef-flash-coremlCore MLfrom Cloudflare/clef-flash—180
JohnP1/d1a-e2b-coremlCore MLfrom google/gemma-4-E2B—168
FluidInference/kev-0.8b-coremlCore MLfrom jaredpalmer/kev-0.8b—147
JohnP1/d1a-e2b-it-qat-coremlCore MLfrom google/gemma-4-E2B-it-qat-w4a16-ct—145
FluidInference/laya-coremlCore MLfrom convaiinnovations/laya—142
FluidInference/d1-omni-600m-coremlCore MLfrom LiquidAI/d1-omni-600M—141
FluidInference/cua-s1-forms-coremlCore MLfrom cua-ai/cua-s1-forms—139
JohnP1/d1a-e2b-it-coreml-4bitCore MLfrom google/gemma-4-E2B-it—139
aac6fef/laya-multilingual-coremlCore MLfrom convaiinnovations/laya-multilingual—118
Pokojaka/modernbert-ja-30m-utterance-taggerCore MLfrom sbintuitions/modernbert-ja-30m—114
Falconsai/intent_classificationCore ML—110
apple/ane-distilbert-base-uncased-finetuned-sst-2-englishCore ML—108
karc14/rv-laya-ft-v2-coremlCore MLfrom convaiinnovations/laya-multilingual—103
JohnP1/d1a-e2b-it-coremlCore MLfrom google/gemma-4-E2B-it—90
aac6fef/laya-multilingual-coreml-aneCore MLfrom convaiinnovations/laya-multilingual—88
spellspeak/audienceCore MLfrom jhu-clsp/ettin-encoder-32m—84
Orvena/Cue-0.4BCore MLfrom Qwen/Qwen3-0.6B-Base—83
pipecat-ai/audienceCore MLfrom jhu-clsp/ettin-encoder-32m—80
faxenoff/code-daemon-relation-v1Core MLfrom hotchpotch/mMiniLMv2-L12-H384—75
JohnP1/d1a-e4b-it-coreml-4bitCore MLfrom google/gemma-4-E4B-it—69
FluidInference/intern-decision-0.8b-coremlCore MLfrom internlm/Intern-Decision-0.8B—63
FluidInference/decision-modernbert-base-coremlCore MLfrom answerdotai/ModernBERT-base—57
NagaYu/crowd-anon-0.1bCore ML—54
aac6fef/laya-multilingual-coreml-adaptive-fp16Core MLfrom convaiinnovations/laya-multilingual—52
aac6fef/laya-coremlCore MLfrom convaiinnovations/laya—51
FluidInference/intern-decision-0.8b-showdown-coremlCore MLfrom internlm/Intern-Decision-0.8B—49
aac6fef/laya-typed-decisions-coremlCore MLfrom convaiinnovations/laya-typed-decisions—47
aac6fef/laya-typed-decisions-coreml-adaptive-fp16Core MLfrom convaiinnovations/laya-typed-decisions—46
faxenoff/code-daemon-denoise-v1Core MLfrom intfloat/multilingual-e5-small—43
aac6fef/laya-coreml-adaptive-fp32Core MLfrom convaiinnovations/laya—41
j-llm/jevliteCore MLfrom answerdotai/ModernBERT-base—39
aac6fef/laya-coreml-adaptive-fp16Core MLfrom convaiinnovations/laya—37
aac6fef/laya-multilingual-coreml-adaptive-fp32Core MLfrom convaiinnovations/laya-multilingual—32
aac6fef/laya-typed-decisions-coreml-adaptive-fp32Core MLfrom convaiinnovations/laya-typed-decisions—32
FluidInference/laya-english-coremlCore MLfrom convaiinnovations/laya—31
FluidInference/gliner2-5-small-coremlCore MLfrom fastino/gliner2.5-small-v1—29
FluidInference/verdict-coremlCore MLfrom heman10x/rlcd-modernbert-151m—29
FluidInference/jeff-coremlCore MLfrom knowledgator/gliformer-large-v1—28
ayjays132/Quantum-NeuralAdaptiveLearningSystemCore ML—26
suryatmodulus/d1-omni-600m-coremlCore MLfrom LiquidAI/d1-omni-600M—26
wiktorwojcik112/ProgrammingLanguageIdentificationCore ML—26
FluidInference/kev-0-5b-coremlCore MLfrom Qwen/Qwen2.5-0.5B—22

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.

Other tasks