Can it run?
Runs well
Using Q4_K_M in llama.cpp
Fit status
Runs well
Decode
128.5 tok/s
TTFT
1507 ms
Safe context
38K
Memory
121.8 GB / 288.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 128.5 tok/s | 2192 ms | 66K |
| Chat | C | Runs well | 128.5 tok/s | 822 ms | 19K |
| Coding | C | Runs well | 128.5 tok/s | 1507 ms | 38K |
| RAG | C | Runs well | 128.5 tok/s | 2740 ms | 66K |
| Reasoning | C | Runs well | 128.5 tok/s | 1781 ms | 38K |
How Mixtral 8x22B (141B params) fits at each quantization level on AMD Instinct MI350X 288GB (288.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 55.0 GB | Low | D34 |
Q3_K_S | 3 | 69.1 GB | Low | D35 |
NVFP4 | 4 |
ollama run mixtral-8x22bhuggingface-cli download mixtral-8x22b79.0 GB |
| Medium |
| D35 |
Q4_K_M | 4 | 86.0 GB | Medium | D36 |
Q5_K_M | 5 | 101.5 GB | High | D37 |
Q6_K | 6 | 115.6 GB | High | D38 |
Q8_0Best for your GPU | 8 | 150.9 GB | Very High | C41 |
F16 | 16 | 289.0 GB | Maximum | F0 |