MacBook Pro M4 32GBBudget pick
C10.1 tok/s decode
~$799 MSRP
Can it run?
Tight fit
Using Q4_K_M in vLLM
Fit status
Tight fit
Decode
24.3 tok/s
TTFT
7973 ms
Safe context
17K
Memory
14.7 GB / 16.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | D | Runs with offload (needs ~0.5 GB host RAM) | 18.7 tok/s | 15091 ms | 30K |
| Chat | C | Tight fit | 19.6 tok/s | 5399 ms | 9K |
| Coding | C | Tight fit | 24.3 tok/s | 7973 ms | 17K |
| RAG | D | Runs with offload (needs ~0.5 GB host RAM) | 18.7 tok/s | 18864 ms | 30K |
| Reasoning | C | Tight fit | 19.6 tok/s | 11698 ms |
How Ministral 3 14B (14B params) fits at each quantization level on RX 7600 XT 16GB (16.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 5.5 GB | Low | D37 |
Q3_K_S | 3 | 6.9 GB | Low | D38 |
NVFP4 | 4 |
huggingface-cli download ministral-3-14bUpgrade options
~$799 MSRP
~$899 MSRP
| 17K |
| Medium |
| D40 |
Q4_K_M | 4 | 8.5 GB | Medium | C41 |
Q5_K_M | 5 | 10.1 GB | High | C43 |
Q6_KBest for your GPU | 6 | 11.5 GB | High | C44 |
Q8_0 | 8 | 15.0 GB | Very High | C44 |
F16 | 16 | 28.7 GB | Maximum | F0 |
~$999 MSRP
RTX 4090 24GBBiggest leap~$1,599 MSRP