Mac Studio M3 Ultra 256GBBudget pick
C101.4 tok/s decode
Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
209.4 tok/s
TTFT
924 ms
Safe context
87K
Memory
17.7 GB / 96.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 244.3 tok/s | 1152 ms | 161K |
| Chat | C | Runs well | 244.3 tok/s | 432 ms | 45K |
| Coding | C | Runs well | 209.4 tok/s | 924 ms | 87K |
| RAG | C | Runs well | 244.3 tok/s | 1441 ms | 161K |
| Reasoning | C | Runs well | 244.3 tok/s | 936 ms | 87K |
How Qwen3.5 9B Uncensored HauhauCS Aggressive (9B params) fits at each quantization level on RTX PRO 6000 Blackwell Server Edition 96GB (96.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 3.5 GB | Low | D30 |
Q3_K_S | 3 | 4.4 GB | Low | D31 |
NVFP4 |
Upgrade options
| 4 |
5.0 GB |
| Medium |
| D31 |
Q4_K_M | 4 | 5.5 GB | Medium | D31 |
Q5_K_M | 5 | 6.5 GB | High | D31 |
Q6_K | 6 | 7.4 GB | High | D31 |
Q8_0 | 8 | 9.6 GB | Very High | D32 |
F16Best for your GPU | 16 | 18.5 GB | Maximum | D34 |