Hardware profile
Mac Studio M4 Max · 64 GB
64 GB system memory · 64 GB unified memory
Platform macOS
Available for models 64 GB
Backends Ollama, llama.cpp, MLX
High memory bandwidth helps with larger models; still reserve memory for the OS and applications.
Model fit
| Model | Minimum | Result |
|---|---|---|
| gpt-oss-20b | 14 GB | Fits |
| Qwen3.6 35B-A3B | 25 GB | Fits |
| Qwen3.8 27B | 20 GB | Fits |
| Gemma 4 12B Unified | 10 GB | Fits |
| Gemma 4 E4B | 8 GB | Fits |
| Gemma 4 31B | 23 GB | Fits |
| gpt-oss-120b | 66 GB | Not recommended |
| DeepSeek V4.1 Flash | 234 GB | Not recommended |
| Llama 3.2 3B | 4 GB | Fits |
| Qwen3 8B | 8 GB | Fits |
| Gemma 3 4B | 6 GB | Fits |
| Phi-4 Mini | 5 GB | Fits |
| Mistral 7B | 8 GB | Fits |
| DeepSeek R1 8B | 8 GB | Fits |