Skip to content

What can the Mac mini M4 Pro 24 GB run?

What fits on the Mac mini M4 Pro 24 GB
Text models: 246Visual models: 12Strongest first
Memory
24 GB unified
Bandwidth
273 GB/s
Maker
Apple

Open the calculator on this machine

Text models that run on it

Text models that run on the Mac mini M4 Pro 24 GB, best first
ModelBest fitContextSpeedStrongest at
Qwen3.8 27B3-bit24K16 tok/sChat, Coding, Agents, Docs
Muse Glimmer 30B3-bit56K15 tok/sDocs
Gemma 4 26B A4B3-bit240K80 tok/sChat, Coding, Agents, Docs
Qwen3.5 9B8-bit152K23 tok/sChat, Coding, Agents, Docs
Gemma 4 12B8-bit256K17 tok/sChat, Coding, Docs
Nemotron 3.5 Lightning 30B A3B3-bit32K71 tok/sAgents, Docs
Qwen3.5 4B8-bit256K40 tok/sAgents, Docs
Devstral Small 2 24B3-bit20K18 tok/sCoding
LFM2.5 24B A2B4-bit56K91 tok/sAgents
Mistral Small 3.2 24B3-bit20K18 tok/sChat, Coding, Agents
Ministral 3 14B6-bit24K19 tok/sCoding, Agents, Docs
Gemma 4 E4B8-bit128K36 tok/sChat, Coding, Docs
gpt-oss 20BMXFP4128K67 tok/sAgents
Ministral 3 8B8-bit40K24 tok/sDocs
Gemma 4 E2B8-bit128K63 tok/sChat
Nemotron 3 Nano 4B8-bit256K48 tok/sDocs
Llama 3.1 8B8-bit52K26 tok/sChat, Coding, Agents, Docs
Qwen3 0.6B8-bit40K172 tok/sNot rated
Qwen3 14B6-bit20K18 tok/sNot rated
Qwen3 VL 8B Instruct8-bit40K24 tok/sText and images
Qwen3 8B8-bit40K26 tok/sNot rated
Qwen2.5 7B Instruct8-bit32K28 tok/sNot rated
Qwen2.5 VL 7B Instruct8-bit112K26 tok/sText and images
Unlimited OCR8-bit32K141 tok/sReading documents
Qwen3 VL 4B Instruct8-bit72K42 tok/sText and images
Qwen3 1.7B8-bit40K84 tok/sNot rated
Qwen3 4B8-bit40K46 tok/sNot rated
Qwen3 4B Instruct 25078-bit72K46 tok/sNot rated
Qwen2.5 3B Instruct8-bit32K57 tok/sNot rated
Qwen3.6 27B3-bit24K16 tok/sText and images
DeepSeek OCR8-bit8K141 tok/sReading documents
Qwen2.5 Coder 14B Instruct6-bit16K18 tok/sCoding
Qwen2.5 VL 3B Instruct8-bit125K48 tok/sText and images
Qwen3.5 27B3-bit24K16 tok/sText and images
NVIDIA Nemotron 3 Nano 30B A3B BF163-bit32K71 tok/sNot rated
Qwen2.5 Coder 7B Instruct8-bit32K28 tok/sCoding
Mistral 7B Instruct V0.28-bit32K28 tok/sNot rated
Chandra OCR 28-bit256K36 tok/sReading documents
Qwen3 VL 2B Instruct8-bit112K81 tok/sText and images
MiniCPM5 2B8-bit128K71 tok/sNot rated
Gemma 3 4B IT8-bit128K50 tok/sText and images
Llama 3.2 3B Instruct8-bit104K57 tok/sNot rated
Qwen2.5 14B Instruct6-bit16K18 tok/sNot rated
Qwen3 1.7B Base8-bit32K96 tok/sNot rated
Diffusiongemma 26B A4B IT3-bit120K77 tok/sText and images
HunyuanOCR8-bit128K138 tok/sText and images
Qwen3 0.6B Base8-bit32K197 tok/sNot rated
Qwen2 VL 7B Instruct8-bit32K26 tok/sText and images
Ornith 1.5 9B8-bit144K23 tok/sNot rated
Unlimited OCR 28-bit128K140 tok/sReading documents
Granite 4.1 3B8-bit128K52 tok/sNot rated
Mistral 7B V0.18-bit32K28 tok/sNot rated
Qwen2.5 Coder 3B Instruct8-bit32K57 tok/sCoding
Meta Llama 3 8B Instruct Abliterated V38-bit8K26 tok/sNot rated
EXAONE 3.5 7.8B Instruct8-bit32K27 tok/sNot rated
Saiga Llama3 8B8-bit8K26 tok/sNot rated
DeepSeek R1 0528 Qwen3 8B8-bit44K26 tok/sNot rated
Gemma 3 12B IT8-bit28K18 tok/sText and images
Phi 3.5 Mini Instruct8-bit28K49 tok/sNot rated
Qwen3 4B Thinking 25078-bit72K46 tok/sNot rated

Best fit is the highest quality, up to 8-bit, that leaves the use case’s context. Ranked by our use-case scores, then by Hugging Face downloads. Speeds are rough, plus or minus 25%. The top 60 of 246 are shown.

Image and video models

Image and video models that run on the Mac mini M4 Pro 24 GB
ModelBest fitPer outputStrongest at
Stable Diffusion XL 1.016-bit80 sImages
Z-Image Turbo8-bit1.9 minImages
FLUX.1 [dev]5-bit4.5 minImages
Qwen-Image 2.14-bit37 minImages
FLUX.3 Klein8-bit2.9 minImages
FLUX.2 [klein] 9B4-bit63 sImages
Krea 2 Turbo5-bit4.4 minImages
Stable Diffusion 3.5 Large5-bit19 minImages
MOSS-Image8-bit21 minImages
Wan 2.2 TI2V-5B8-bit3.0 hVideo
MOSS Image Base8-bit21 minImages
HunyuanVideo 1.54-bit15.5 hVideo

Only just too big

Questions

What is the best LLM for the Mac mini M4 Pro 24 GB?

Of the open models we track, the strongest that run well on the Mac mini M4 Pro 24 GB are Qwen3.8 27B, Muse Glimmer 30B, Gemma 4 26B A4B. 246 text models fit it in all.

Calculated from published specs, not benchmarked. All machines