Mac Studio M2 Ultra 128GBBudget pick
C95.1 tok/s decode
~$3,999 MSRP
Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
285.5 tok/s
TTFT
678 ms
Safe context
83K
Memory
15.3 GB / 80.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 309.3 tok/s | 910 ms | 154K |
| Chat | C | Runs well | 309.3 tok/s | 350 ms | 43K |
| Coding | C | Runs well | 285.5 tok/s | 678 ms | 83K |
| RAG | C | Runs well | 309.3 tok/s | 1138 ms | 154K |
| Reasoning | C | Runs well | 309.3 tok/s | 740 ms | 83K |
How DeepSeek R1 0528 Qwen3 8B (8B params) fits at each quantization level on NVIDIA A800 80GB (80.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 3.1 GB | Low | D30 |
Q3_K_S | 3 | 3.9 GB | Low | D30 |
NVFP4 | 4 |
Upgrade options
~$3,999 MSRP
~$3,999 MSRP
4.5 GB |
| Medium |
| D31 |
Q4_K_M | 4 | 4.9 GB | Medium | D31 |
Q5_K_M | 5 | 5.8 GB | High | D31 |
Q6_K | 6 | 6.6 GB | High | D31 |
Q8_0 | 8 | 8.6 GB | Very High | D32 |
F16Best for your GPU | 16 | 16.4 GB | Maximum | D34 |