AI model library, optimized for on-device inference.

Qwen 3.5

From Alibaba

8 checkpoints

0.8BQwen 3.5

Speed519 tok/sSize460 MB

Speed392 tok/sSize824 MB

2BQwen 3.5

Speed319 tok/sSize1.1 GB

Speed216 tok/sSize2.0 GB

4BQwen 3.5

Speed159 tok/sSize2.5 GB

Speed101 tok/sSize4.5 GB

9BQwen 3.5

Speed101 tok/sSize5.2 GB

Speed63 tok/sSize9.3 GB

Qwen 3.6

From Alibaba

2 checkpoints

27BQwen 3.6

Speed119 tok/sSize15.6 GB

Speed82 tok/sSize28.8 GB

Qwen 3.8

From Alibaba

2 checkpoints

27BQwen 3.8

Speed36 tok/sSize15.5 GB

Speed20 tok/sSize28.8 GB

LFM 2.5

From LiquidAI

10 checkpoints

230MLFM 2.5

Speed1392 tok/sSize130 MB

Speed1110 tok/sSize243 MB

350MLFM 2.5

Speed1116 tok/sSize198 MB

Speed829 tok/sSize373 MB

1.2B InstructLFM 2.5

Speed583 tok/sSize639 MB

Speed377 tok/sSize1.2 GB

1.2B ThinkingLFM 2.5

Speed578 tok/sSize639 MB

Speed372 tok/sSize1.2 GB

2.6BLFM 2.5

Speed282 tok/sSize1.5 GB

Speed171 tok/sSize2.8 GB

Muse-Glimmer

From Meta

2 checkpoints

30BMuse-Glimmer

Speed33 tok/sSize16.5 GB

Speed19 tok/sSize30.2 GB