AI models library optimized for on device.

Search our library of available models to run on device and deploy in seconds with Mirai.

Our quantized models:

trymirai/Qwen3.5-0.8B-M
438.0 MB800M
trymirai/Qwen3.5-0.8B-L
785.4 MB800M
trymirai/Qwen3.5-2B-M
1.0 GB2B
trymirai/Qwen3.5-2B-L
1.9 GB2B
trymirai/Qwen3.5-4B-M
2.3 GB4B
trymirai/Qwen3.5-4B-L
4.2 GB4B

Ecosystem partner models:

LiquidAI/LFM2.5-350M
680.8 MB350M
LiquidAI/LFM2.5-1.2B-Instruct
2.2 GB1.2B
LiquidAI/LFM2.5-1.2B-Thinking
2.2 GB1.2B
LiquidAI/LFM2.5-1.2B-Instruct-MLX-4bit
632.8 MB1.2B
mlx-community/LFM2.5-1.2B-Thinking-4bit
632.8 MB1.2B
LiquidAI/LFM2.5-1.2B-Instruct-MLX-8bit
1.2 GB1.2B
mlx-community/LFM2.5-1.2B-Thinking-8bit
1.2 GB1.2B

All supported models:

google/functiongemma-270m-it
543.4 MB270M
google/gemma-3-1b-it
1.9 GB1B
mlx-community/gemma-3-1b-it-4bit
568.8 MB1B
mlx-community/gemma-3-1b-it-8bit
1.0 GB1B
google/gemma-3-4b-it
7.3 GB4B
mlx-community/gemma-3-4b-it-4bit
2.1 GB4B
mlx-community/gemma-3-4b-it-8bit
3.9 GB4B
google/gemma-3-27b-it
50.3 GB27B
mlx-community/gemma-3-27b-it-4bit
14.2 GB27B
mlx-community/gemma-3-27b-it-8bit
26.8 GB27B
openai/gpt-oss-20b
39.0 GB20B
LiquidAI/LFM2-350M
680.8 MB350M
mlx-community/LFM2-350M-4bit
195.0 MB350M
mlx-community/LFM2-350M-8bit
364.0 MB350M
LiquidAI/LFM2-700M
1.4 GB700M
mlx-community/LFM2-700M-4bit
403.2 MB700M
mlx-community/LFM2-700M-8bit
757.2 MB700M
LiquidAI/LFM2-1.2B
2.2 GB1.2B
mlx-community/LFM2-1.2B-4bit
632.8 MB1.2B
mlx-community/LFM2-1.2B-8bit
1.2 GB1.2B
LiquidAI/LFM2-2.6B
4.8 GB2.6B
mlx-community/LFM2-2.6B-4bit
1.4 GB2.6B
mlx-community/LFM2-2.6B-8bit
2.5 GB2.6B
meta-llama/Llama-3.2-1B-Instruct
2.3 GB1B
mlx-community/Llama-3.2-1B-Instruct-4bit
679.7 MB1B
mlx-community/Llama-3.2-1B-Instruct-8bit
1.2 GB1B
meta-llama/Llama-3.2-3B-Instruct
6.0 GB3B
mlx-community/Llama-3.2-3B-Instruct-4bit
1.7 GB3B
mlx-community/Llama-3.2-3B-Instruct-8bit
3.2 GB3B
Qwen/Qwen3-0.6B
1.1 GB600M
Qwen/Qwen3-0.6B-MLX-4bit
313.3 MB600M
Qwen/Qwen3-0.6B-MLX-8bit
597.4 MB600M
Qwen/Qwen3-1.7B
3.2 GB1.7B
Qwen/Qwen3-1.7B-MLX-4bit
883.2 MB1.7B
Qwen/Qwen3-1.7B-MLX-8bit
1.7 GB1.7B
Qwen/Qwen3-4B
7.5 GB4B
Qwen/Qwen3-4B-Instruct-2507
7.5 GB4B
Qwen/Qwen3-4B-Thinking-2507
7.5 GB4B
Qwen/Qwen3-4B-MLX-4bit
2.0 GB4B
Qwen/Qwen3-4B-MLX-8bit
3.9 GB4B
Qwen/Qwen3-8B
15.3 GB8B
Qwen/Qwen3-8B-MLX-4bit
4.1 GB8B
Qwen/Qwen3-8B-MLX-8bit
7.9 GB8B
Qwen/Qwen3-14B
27.5 GB14B
Qwen/Qwen3-14B-MLX-4bit
7.3 GB14B
Qwen/Qwen3-14B-MLX-8bit
14.2 GB14B
Qwen/Qwen3-32B
61.0 GB32B
Qwen/Qwen3-32B-MLX-4bit
16.2 GB32B
Qwen/Qwen3-32B-MLX-8bit
31.5 GB32B
Qwen/Qwen3.5-0.8B
1.4 GB800M
mlx-community/Qwen3.5-0.8B-MLX-4bit
424.5 MB800M
mlx-community/Qwen3.5-0.8B-MLX-8bit
783.0 MB800M
Qwen/Qwen3.5-2B
3.5 GB2B
mlx-community/Qwen3.5-2B-MLX-4bit
1.0 GB2B
mlx-community/Qwen3.5-2B-MLX-8bit
1.9 GB2B
Qwen/Qwen3.5-4B
7.9 GB4B
mlx-community/Qwen3.5-4B-MLX-4bit
2.2 GB4B
mlx-community/Qwen3.5-4B-MLX-8bit
4.2 GB4B
Qwen/Qwen3.5-9B
16.7 GB9B
mlx-community/Qwen3.5-9B-MLX-4bit
4.7 GB9B
mlx-community/Qwen3.5-9B-MLX-8bit
8.9 GB9B
Qwen/Qwen3.5-27B
50.1 GB27B
mlx-community/Qwen3.5-27B-4bit
14.1 GB27B
mlx-community/Qwen3.5-27B-8bit
26.6 GB27B
Qwen/Qwen3.6-27B
53.5 GB27B