Mac mini M4 64GBBudget pick
C4.4 tok/s decode
~$1,099 MSRP
Can it run?
Very compromised
Using Q4_K_M in Ollama
Fit status
Very compromised
Decode
4.8 tok/s
TTFT
40264 ms
Safe context
14K
Memory
29.6 GB / 25.9 GB
Offload
10%
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | F | Too heavy | 5.6 tok/s | 50199 ms | 24K |
| Chat | C | Runs with offload (needs ~0.9 GB host RAM) | 5.5 tok/s | 19076 ms | 8K |
| Coding | D | Very compromised | 4.8 tok/s | 40264 ms | 14K |
| RAG | F | Too heavy | 5.6 tok/s | 62749 ms | 24K |
| Reasoning | D | Very compromised (needs ~2.4 GB host RAM) | 5.4 tok/s | 42370 ms |
How Qwen 2.5 Coder 32B (32B params) fits at each quantization level on MacBook Pro M3 Pro 36GB (25.9 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 12.5 GB | Low | D40 |
Q3_K_S | 3 | 15.7 GB | Low | C42 |
NVFP4 | 4 |
ollama run qwen-2.5-coder-32bhuggingface-cli download qwen-2.5-coder-32bUpgrade options
~$1,099 MSRP
~$1,599 MSRP
| 14K |
17.9 GB |
| Medium |
| C44 |
Q4_K_MBest for your GPU | 4 | 19.5 GB | Medium | C44 |
Q5_K_M | 5 | 23.0 GB | High | C44 |
Q6_K | 6 | 26.2 GB | High | F0 |
Q8_0 | 8 | 34.2 GB | Very High | F0 |
F16 | 16 | 65.6 GB | Maximum | F0 |
~$3,999 MSRP
NVIDIA A100 40GBBiggest leap~$10,000 MSRP