Skip to content
Apple · on-device

Apple Core AI models

The models Apple supports converting to its Core AI .aimodel format for running on-device: 58 models across 70 export presets, read directly from apple/coreai-models, checked 19m ago. Apple publishes the export recipes there, not the converted models: each entry tells you which model, which platforms, and the settings Apple uses.

Latest repository change 3h ago: Route reasoning to reasoning_content when the chat template opens the think block (#508)

Alert me about new Apple/CoreAI models

Clear filters

Language models 12

gemma-4-E2B-it

google/gemma-4-E2B-it
iOS
Google · Gemma 1 preset
Family
gemma4
Context
128K
Compression
none

gemma-4-E4B-it

google/gemma-4-E4B-it
iOS
Google · Gemma 1 preset
Family
gemma4
Context
128K
Compression
none

Mistral-7B-Instruct-v0.3

mistralai/Mistral-7B-Instruct-v0.3
macOSiOS
Mistral 2 presets
Family
mistral
Context
8K
Compression
4bit, 4bit_weight_palettized_group8

OLMo-2-0425-1B-Instruct

allenai/OLMo-2-0425-1B-Instruct
macOSiOS
Ai2 · OLMo 2 presets
Family
olmo2
Context
4K
Compression
4bit, none

Qwen2.5-1.5B-Instruct

Qwen/Qwen2.5-1.5B-Instruct
macOSiOS
Qwen 3 presets
Family
qwen2.5
Context
32K
Compression
4bit, 4bit_weights_8bit_kv_cache, 4bit_weight_palettized_group8

Qwen3-0.6B

Qwen/Qwen3-0.6B
macOSiOS
Qwen 2 presets
Family
qwen3
Context
8K
Compression
4bit, none

Qwen3-1.7B

Qwen/Qwen3-1.7B
macOSiOS
Qwen 2 presets
Family
qwen3
Context
32K
Compression
4bit, none

Qwen3-4B

Qwen/Qwen3-4B
macOSiOS
Qwen 3 presets
Family
qwen3
Context
40K
Compression
4bit, 4bit_weights_8bit_kv_cache, none

Qwen3-8B

Qwen/Qwen3-8B
macOSiOS
Qwen 2 presets
Family
qwen3
Context
40K
Compression
4bit, none

SmolLM2-1.7B-Instruct

HuggingFaceTB/SmolLM2-1.7B-Instruct
macOSiOS
Hugging Face 2 presets
Family
smollm2
Context
8K
Compression
4bit, none

SmolLM2-135M-Instruct

HuggingFaceTB/SmolLM2-135M-Instruct
macOSiOS
Hugging Face 2 presets
Family
smollm2
Context
8K
Compression
none

SmolLM2-360M-Instruct

HuggingFaceTB/SmolLM2-360M-Instruct
macOSiOS
Hugging Face 2 presets
Family
smollm2
Context
8K
Compression
none

Embeddings 3

clap-htsat-unfused

laion/clap-htsat-unfused
macOSiOS
laion 1 preset
Family
clap
Context
—
Compression
—

clip-vit-base-patch32

openai/clip-vit-base-patch32
macOSiOS
OpenAI · gpt-oss 1 preset
Family
clip
Context
—
Compression
—

embeddinggemma-2

google/embeddinggemma-2
macOSiOS
Google · Gemma 1 preset
Family
embeddinggemma
Context
—
Compression
—

Speech recognition 3

wav2vec2-base

wav2vec2_asr_base_960h
macOSiOS
wav2vec2 1 preset
Family
wav2vec2
Context
—
Compression
—

whisper-large-v3

openai/whisper-large-v3
macOSiOS
OpenAI · gpt-oss 1 preset
Family
whisper
Context
—
Compression
—

whisper-large-v3-turbo

openai/whisper-large-v3-turbo
macOSiOS
OpenAI · gpt-oss 1 preset
Family
whisper
Context
—
Compression
—

Object detection 2

yolos-base

hustvl/yolos-base
macOSiOS
hustvl 1 preset
Family
yolo
Context
—
Compression
—

yolos-tiny

hustvl/yolos-tiny
macOSiOS
hustvl 1 preset
Family
yolo
Context
—
Compression
—

Segmentation 3

efficient-sam-vitt

efficient_sam_vitt
macOSiOS
efficient-sam 1 preset
Family
efficient-sam
Context
—
Compression
—

sam3

facebook/sam3
macOSiOS
facebook 1 preset
Family
sam3
Context
—
Compression
—

sam3

facebook/sam3
macOSiOS
facebook 1 preset
Family
sam3_video
Context
—
Compression
—

Super-resolution 1

edsr-x2

edsr_r16f64_x2
macOSiOS
edsr 1 preset
Family
edsr
Context
—
Compression
—

Text encoders 4

roberta-base

roberta-base
macOSiOS
roberta 1 preset
Family
roberta
Context
—
Compression
—

t5-base

google-t5/t5-base
macOSiOS
google-t5 1 preset
Family
t5
Context
—
Compression
—

t5-large

google-t5/t5-large
macOSiOS
google-t5 1 preset
Family
t5
Context
—
Compression
—

t5-small

google-t5/t5-small
macOSiOS
google-t5 1 preset
Family
t5
Context
—
Compression
—

Image classification 1

pvt-v2-b0

pvt_v2_b0
macOSiOS
pvt 1 preset
Family
pvt
Context
—
Compression
—