Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
105.9 tok/s
TTFT
1829 ms
Safe context
50K
Memory
8.3 GB / 25.9 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 105.9 tok/s | 2660 ms | 94K |
| Chat | C | Runs well | 105.9 tok/s | 998 ms | 25K |
| Coding | C | Runs well | 105.9 tok/s | 1829 ms | 50K |
| RAG | C | Runs well | 105.9 tok/s | 3325 ms | 94K |
| Reasoning | C | Runs well | 105.9 tok/s | 2161 ms | 50K |
How Qwen3.5 4B (4B params) fits at each quantization level on MacBook Pro M4 Max 36GB (25.9 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 1.6 GB | Low | D31 |
Q3_K_S | 3 | 2.0 GB | Low | D31 |
NVFP4 | 4 |
2.2 GB |
| Medium |
| D31 |
Q4_K_M | 4 | 2.4 GB | Medium | D31 |
Q5_K_M | 5 | 2.9 GB | High | D32 |
Q6_K | 6 | 3.3 GB | High | D32 |
Q8_0 | 8 | 4.3 GB | Very High | D33 |
F16Best for your GPU | 16 | 8.2 GB | Maximum | D36 |