AMD Instinct MI350X 288GBBudget pick
C86.3 tok/s decode
~$8,000 MSRP
Can it run?
Very compromised
Using Q4_K_M in vLLM
Fit status
Very compromised
Decode
7.5 tok/s
TTFT
25909 ms
Safe context
15K
Memory
101.3 GB / 92.2 GB
Offload
10%
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | F | Too heavy | 6.5 tok/s | 43334 ms | 25K |
| Chat | C | Runs with offload (needs ~0.3 GB host RAM) | 6.5 tok/s | 16274 ms | 8K |
| Coding | D | Very compromised | 7.5 tok/s | 25909 ms | 15K |
| RAG | F | Too heavy | 6.5 tok/s | 54168 ms | 25K |
| Reasoning | D | Very compromised (needs ~6.1 GB host RAM) | 6.3 tok/s | 36187 ms |
How Command A 111B (111B params) fits at each quantization level on Mac Studio M1 Ultra 128GB (92.2 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 43.3 GB | Low | D39 |
Q3_K_S | 3 | 54.4 GB | Low | C42 |
NVFP4 | 4 |
ollama run command-a-111bhuggingface-cli download command-a-111bUpgrade options
~$8,000 MSRP
~$15,000 MSRP
| 15K |
62.2 GB |
| Medium |
| C44 |
Q4_K_MBest for your GPU | 4 | 67.7 GB | Medium | C44 |
Q5_K_M | 5 | 79.9 GB | High | C44 |
Q6_K | 6 | 91.0 GB | High | C44 |
Q8_0 | 8 | 118.8 GB | Very High | F0 |
F16 | 16 | 227.6 GB | Maximum | F0 |
~$30,000 MSRP