MacBook Pro M4 Max 96GBBudget pick
C16.6 tok/s decode
~$2,499 MSRP
Can it run?
Runs well
Using Q4_K_M in llama.cpp
Fit status
Runs well
Decode
3.8 tok/s
TTFT
51478 ms
Safe context
22K
Memory
33.9 GB / 46.1 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Tight fit | 4.2 tok/s | 67597 ms | 38K |
| Chat | C | Runs well | 4.2 tok/s | 25349 ms | 12K |
| Coding | C | Runs well | 3.8 tok/s | 51478 ms | 22K |
| RAG | C | Tight fit | 4.2 tok/s | 84496 ms | 38K |
| Reasoning | C | Runs well | 4.2 tok/s | 54923 ms | 22K |
How Yi 34B Chat (34B params) fits at each quantization level on Mac mini M4 64GB (46.1 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 13.3 GB | Low | D35 |
Q3_K_S | 3 | 16.7 GB | Low | D37 |
NVFP4 | 4 |
ollama run yi-34b-chathuggingface-cli download yi-34b-chatUpgrade options
~$2,499 MSRP
RTX A6000 48GBBest value~$4,650 MSRP
RTX PRO 5000 Blackwell 48GBBiggest leap
19.0 GB |
| Medium |
| D38 |
Q4_K_M | 4 | 20.7 GB | Medium | D39 |
Q5_K_M | 5 | 24.5 GB | High | C41 |
Q6_K | 6 | 27.9 GB | High | C42 |
Q8_0Best for your GPU | 8 | 36.4 GB | Very High | C44 |
F16 | 16 | 69.7 GB | Maximum | F0 |