Models in Apple’s Core AI catalogue. Apple publishes export recipes in apple/coreai-models that convert each one to run on-device.
Qwen3.8-27B
Downloads, 30 days
169K
Memory needed (est.)
3 GB
Quantisation
4-bit, MXFP4, 8-bit +2
Devices
Runs on: Mac
Fits in unified memory
8
16
24
32
64
128
apache-2.0Added 14 Aug 2026
Qwen3.6-35B-A3B
Downloads, 30 days
57K
Memory needed (est.)
3 GB
Quantisation
4-bit, MXFP4, 5-bit +4
Devices
Runs on: Mac
Fits in unified memory
8
16
24
32
64
128
apache-2.0Added 16 Apr 2026
Qwen3.6-27B
Downloads, 30 days
55K
Memory needed (est.)
3 GB
Quantisation
4-bit, MXFP4, 5-bit +4
Devices
Runs on: Mac
Fits in unified memory
8
16
24
32
64
128
apache-2.0Added 22 Apr 2026
gemma-4-E4B-it
Downloads, 30 days
31K
Memory needed (est.)
8 GB
Quantisation
4-bit, MXFP4, 5-bit +4
Devices
Runs on: Mac, iPhone
Fits in unified memory
8
16
24
32
64
128
apache-2.0Added 2 Apr 2026
gemma-4-E2B-it
Downloads, 30 days
29K
Memory needed (est.)
5 GB
Quantisation
4-bit, MXFP4, 5-bit +4
Devices
Runs on: Mac, iPhone
Fits in unified memory
8
16
24
32
64
128
apache-2.0Added 2 Apr 2026
gemma-3n-E4B
Downloads, 30 days
749
Memory needed (est.)
8 GB
Quantisation
3-bit, 4-bit, 5-bit +3
Devices
Runs on: Mac
Fits in unified memory
8
16
24
32
64
128
gemmaAdded 26 Jun 2025
sam3
Downloads, 30 days
584
Memory needed (est.)
3 GB
Quantisation
4-bit, MXFP4, 5-bit +4
Devices
Runs on: Mac, iPhone
Fits in unified memory
8
16
24
32
64
128
otherAdded 24 Dec 2025
gemma-3-4b-it
Downloads, 30 days
297
Memory needed (est.)
5 GB
Quantisation
3-bit, 4-bit DWQ, 5-bit +3
Devices
Runs on: Mac
Fits in unified memory
8
16
24
32
64
128
gemmaAdded 15 Apr 2025
Showing 8 of 8