AI models library optimized for on device.

Qwen models from Alibaba

Qwen3.827BQwen3.627BQwen3.59BQwen3.54BQwen3.52BQwen3.50.8B

Qwen3.8 27B

2 available checkpoints

MediumQuantization129 tok/sSpeed

Measured on Apple M5 Max 128GB with speculative decoding

15.7 GBSize

LargeQuantization77 tok/sSpeed

Measured on Apple M5 Max 128GB with speculative decoding

28.9 GBSize

LFM models from LiquidAI

LFM2.52.6BLFM2.51.2B ThinkingLFM2.51.2B InstructLFM2.5350MLFM2.5230M

LFM2.5 2.6B

2 available checkpoints

MediumQuantization267 tok/sSpeed

Measured on Apple M5 Max 128GB

1.5 GBSize

LargeQuantization175 tok/sSpeed

Measured on Apple M5 Max 128GB

2.8 GBSize

Muse Glimmer models from Meta

Muse-Glimmer 30B

2 available checkpoints

MediumQuantization90 tok/sSpeed

Measured on Apple M5 Max 128GB with speculative decoding

16.6 GBSize

LargeQuantization65 tok/sSpeed

Measured on Apple M5 Max 128GB with speculative decoding

30.3 GBSize