Can it run?
Runs well
Using Q4_K_M in vLLM
Fit status
Runs well
Decode
61.1 tok/s
TTFT
3171 ms
Safe context
29K
Memory
106.7 GB / 192.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 61.1 tok/s | 4612 ms | 50K |
| Chat | C | Runs well | 61.1 tok/s | 1730 ms | 16K |
| Coding | C | Runs well | 61.1 tok/s | 3171 ms | 29K |
| RAG | C | Runs well | 61.1 tok/s | 5765 ms | 50K |
| Reasoning | C | Runs well | 61.1 tok/s | 3747 ms | 29K |
How Command A 111B (111B params) fits at each quantization level on AMD Instinct MI300X 192GB (192.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 43.3 GB | Low | D34 |
Q3_K_S | 3 | 54.4 GB | Low | D35 |
NVFP4 | 4 | 62.2 GB | Medium | D36 |
Q4_K_M | 4 | 67.7 GB | Medium | D37 |
Q5_K_M | 5 | 79.9 GB | High | D38 |
Q6_K | 6 | 91.0 GB | High | D40 |
Q8_0Best for your GPU | 8 | 118.8 GB | Very High | C43 |
F16 | 16 | 227.6 GB | Maximum | F0 |
ollama run command-a-111bhuggingface-cli download command-a-111b