MacBook Pro M4 32GBBudget pick
C9.6 tok/s decode
~$799 MSRP
Can it run?
Tight fit
Using Q4_K_M in Ollama
Fit status
Tight fit
Decode
23.1 tok/s
TTFT
8372 ms
Safe context
18K
Memory
14.1 GB / 16.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs with offload (needs ~0.2 GB host RAM) | 18.3 tok/s | 15406 ms | 31K |
| Chat | C | Runs well | 18.6 tok/s | 5669 ms | 10K |
| Coding | C | Tight fit | 23.1 tok/s | 8372 ms | 18K |
| RAG | C | Runs with offload (needs ~0.2 GB host RAM) | 18.3 tok/s | 19257 ms | 31K |
| Reasoning | C | Tight fit | 18.6 tok/s | 12282 ms |
How Phi-4-reasoning-plus 14B (14.699999809265137B params) fits at each quantization level on RX 7600 XT 16GB (16.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 5.7 GB | Low | D37 |
Q3_K_S | 3 | 7.2 GB | Low | D39 |
NVFP4 | 4 |
ollama run phi-4-reasoning-plus-14bhuggingface-cli download phi-4-reasoning-plus-14bUpgrade options
~$799 MSRP
~$899 MSRP
| 18K |
| Medium |
| C40 |
Q4_K_M | 4 | 9.0 GB | Medium | C41 |
Q5_K_M | 5 | 10.6 GB | High | C43 |
Q6_KBest for your GPU | 6 | 12.1 GB | High | C44 |
Q8_0 | 8 | 15.7 GB | Very High | C44 |
F16 | 16 | 30.1 GB | Maximum | F0 |
~$999 MSRP
RTX A4500 20GBBiggest leap