What can the Mac mini M4 32 GB run?
| Text models: 265 | Visual models: 16 | Strongest first |
- Memory
- 32 GB unified
- Bandwidth
- 120 GB/s
- Maker
- Apple
Text models that run on it
| Model | Best fit | Context | Speed | Strongest at |
|---|---|---|---|---|
| Qwen3.8 27B | 4-bit | 56K | 5.9 tok/s | Chat, Coding, Agents, Docs |
| Muse Glimmer 30B | 4-bit | 128K | 5.6 tok/s | Docs |
| Qwen3.6 35B A3B | 3-bit | 152K | 47 tok/s | Chat, Coding, Agents, Docs |
| Gemma 4 31B | 4-bit | 20K | 5.1 tok/s | Chat, Coding, Agents, Docs |
| Gemma 4 26B A4B | 5-bit | 216K | 32 tok/s | Chat, Coding, Agents, Docs |
| Qwen3.5 9B | 8-bit | 256K | 10 tok/s | Chat, Coding, Agents, Docs |
| GLM-4.7 Flash | 4-bit | 36K | 40 tok/s | Chat, Coding, Agents |
| Gemma 4 12B | 8-bit | 256K | 7.5 tok/s | Chat, Coding, Docs |
| Nemotron 3.5 Lightning 30B A3B | 4-bit | 256K | 38 tok/s | Agents, Docs |
| Qwen3.5 4B | 8-bit | 256K | 19 tok/s | Agents, Docs |
| Devstral Small 2 24B | 5-bit | 20K | 5.8 tok/s | Coding |
| LFM2.5 24B A2B | 6-bit | 64K | 45 tok/s | Agents |
| Mistral Small 3.2 24B | 5-bit | 20K | 5.8 tok/s | Chat, Coding, Agents |
| Ministral 3 14B | 8-bit | 36K | 6.8 tok/s | Coding, Agents, Docs |
| Gemma 4 E4B | 8-bit | 128K | 17 tok/s | Chat, Coding, Docs |
| gpt-oss 20B | MXFP4 | 128K | 34 tok/s | Agents |
| Ministral 3 8B | 8-bit | 80K | 11 tok/s | Docs |
| Gemma 4 E2B | 8-bit | 128K | 31 tok/s | Chat |
| Olmo 3.1 32B | 3-bit | 64K | 6.1 tok/s | Chat, Coding, Agents, Docs |
| Llama 3.1 8B | 8-bit | 88K | 12 tok/s | Chat, Coding, Agents, Docs |
| Qwen3 14B | 8-bit | 28K | 6.5 tok/s | Not rated |
| Qwen3 VL 8B Instruct | 8-bit | 72K | 11 tok/s | Text and images |
| Qwen3 8B | 8-bit | 40K | 12 tok/s | Not rated |
| Qwen2.5 7B Instruct | 8-bit | 32K | 13 tok/s | Not rated |
| Qwen2.5 VL 7B Instruct | 8-bit | 125K | 12 tok/s | Text and images |
| Unlimited OCR | 8-bit | 32K | 71 tok/s | Reading documents |
| Qwen3 VL 4B Instruct | 8-bit | 104K | 20 tok/s | Text and images |
| Qwen3 1.7B | 8-bit | 40K | 41 tok/s | Not rated |
| Qwen3 4B | 8-bit | 40K | 22 tok/s | Not rated |
| Qwen3 32B | 3-bit | 16K | 6.1 tok/s | Not rated |
| Qwen3 4B Instruct 2507 | 8-bit | 104K | 22 tok/s | Not rated |
| Qwen3.6 27B | 4-bit | 56K | 5.9 tok/s | Text and images |
| Qwen2.5 Coder 14B Instruct | 8-bit | 24K | 6.5 tok/s | Coding |
| Qwen2.5 VL 3B Instruct | 8-bit | 125K | 23 tok/s | Text and images |
| Qwen3 Coder 30B A3B Instruct | 4-bit | 24K | 40 tok/s | Coding |
| Qwen3.5 27B | 4-bit | 56K | 5.9 tok/s | Text and images |
| Qwen3.5 35B A3B | 3-bit | 152K | 47 tok/s | Text and images |
| NVIDIA Nemotron 3 Nano 30B A3B BF16 | 4-bit | 256K | 38 tok/s | Not rated |
| Qwen2.5 Coder 7B Instruct | 8-bit | 32K | 13 tok/s | Coding |
| Mistral 7B Instruct V0.2 | 8-bit | 32K | 13 tok/s | Not rated |
| Chandra OCR 2 | 8-bit | 256K | 17 tok/s | Reading documents |
| Qwen3 30B A3B | 4-bit | 24K | 40 tok/s | Not rated |
| Qwen3 VL 2B Instruct | 8-bit | 160K | 40 tok/s | Text and images |
| Gemma 3 4B IT | 8-bit | 128K | 24 tok/s | Text and images |
| Qwen3 30B A3B Instruct 2507 | 4-bit | 24K | 40 tok/s | Not rated |
| Llama 3.2 3B Instruct | 8-bit | 128K | 27 tok/s | Not rated |
| Laguna XS.2 | 4-bit | 20K | 41 tok/s | Not rated |
| OTel 2.0 LLM 31B IT | 4-bit | 20K | 5.1 tok/s | Not rated |
| Qwen2.5 32B Instruct | 3-bit | 16K | 6.1 tok/s | Not rated |
| MiniCPM V 4.7 35B A3B | 3-bit | 168K | 47 tok/s | Text and images |
| Qwen2.5 14B Instruct | 8-bit | 24K | 6.5 tok/s | Not rated |
| Laguna XS 2.1 | 4-bit | 20K | 41 tok/s | Not rated |
| Qwen3 1.7B Base | 8-bit | 32K | 48 tok/s | Not rated |
| Diffusiongemma 26B A4B IT | 5-bit | 104K | 30 tok/s | Text and images |
| Qwen2.5 VL 32B Instruct | 3-bit | 16K | 6.0 tok/s | Text and images |
| Qwen2.5 Coder 32B Instruct | 3-bit | 16K | 6.1 tok/s | Coding |
| Qwen2 VL 7B Instruct | 8-bit | 32K | 12 tok/s | Text and images |
| Ornith 1.5 9B | 8-bit | 256K | 10 tok/s | Not rated |
| Unlimited OCR 2 | 8-bit | 128K | 71 tok/s | Reading documents |
| Granite 4.1 3B | 8-bit | 128K | 25 tok/s | Not 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 265 are shown.
Image and video models
| Model | Best fit | Per output | Strongest at |
|---|---|---|---|
| Stable Diffusion XL 1.0 | 16-bit | 2.7 min | Images |
| Z-Image Turbo | 16-bit | 3.9 min | Images |
| FLUX.1 [dev] | 8-bit | 7.8 min | Images |
| Qwen-Image 2.1 | 8-bit | 64 min | Images |
| FLUX.3 Klein | 16-bit | 5.7 min | Images |
| Wan2.2 I2V A14B | 4-bit | 25.4 h | Video |
| FLUX.2 [klein] 9B | 8-bit | 1.8 min | Images |
| Wan2.1 I2V 14B | 4-bit | 25.4 h | Video |
| Wan 2.1 T2V-14B | 4-bit | 31.7 h | Video |
| Krea 2 Turbo | 8-bit | 7.7 min | Images |
| Stable Diffusion 3.5 Large | 8-bit | 34 min | Images |
| MOSS-Image | 16-bit | 42 min | Images |
| Wan 2.2 TI2V-5B | 16-bit | 6.0 h | Video |
| MOSS Image Base | 16-bit | 42 min | Images |
| Wan 2.2 A14B | 4-bit | 25.4 h | Video |
| HunyuanVideo 1.5 | 8-bit | 26.9 h | Video |
Only just too big
- Qwen-Image
- Kimi Linear 48B A3B Instruct
- Qwen Image 2512
- Seed OSS 36B Instruct
- Phi 3.5 MOE Instruct
- K2 Horizon MoVA 36B A4B
These need up to a quarter more memory than the Mac mini M4 32 GB has. Spilling into system RAM or a lower precision can get them running, slowly.
Questions
What is the best LLM for the Mac mini M4 32 GB?
Of the open models we track, the strongest that run well on the Mac mini M4 32 GB are Qwen3.8 27B, Muse Glimmer 30B, Qwen3.6 35B A3B. 265 text models fit it in all.
Calculated from published specs, not benchmarked. All machines