AMD Instinct MI350X 288GBBudget pick
C86.3 tok/s decode
~$8,000 MSRP
Can it run?
Runs well
Using Q4_K_M in vLLM
Fit status
Runs well
Decode
64.7 tok/s
TTFT
2993 ms
Safe context
36K
Memory
113.1 GB / 256.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 64.7 tok/s | 4353 ms | 63K |
| Chat | C | Runs well | 64.7 tok/s | 1632 ms | 20K |
| Coding | C | Runs well | 64.7 tok/s | 2993 ms | 36K |
| RAG | C | Runs well | 64.7 tok/s | 5442 ms | 63K |
| Reasoning | C | Runs well | 64.7 tok/s | 3537 ms | 36K |
How Command A 111B (111B params) fits at each quantization level on AMD Instinct MI325X 256GB (256.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 43.3 GB | Low | D33 |
Q3_K_S | 3 | 54.4 GB | Low | D34 |
NVFP4 | 4 | 62.2 GB | Medium | D35 |
Q4_K_M | 4 | 67.7 GB | Medium | D35 |
Q5_K_M | 5 | 79.9 GB | High | D36 |
Q6_K | 6 | 91.0 GB | High | D37 |
Q8_0Best for your GPU | 8 | 118.8 GB | Very High | D39 |
F16 | 16 | 227.6 GB | Maximum | C44 |
ollama run command-a-111bhuggingface-cli download command-a-111bUpgrade options
~$8,000 MSRP