Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
46.2 tok/s
TTFT
4186 ms
Safe context
21K
Memory
36.7 GB / 48.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 46.2 tok/s | 6089 ms | 33K |
| Chat | C | Runs well | 46.2 tok/s | 2283 ms | 11K |
| Coding | C | Runs well | 46.2 tok/s | 4186 ms | 21K |
| RAG | C | Runs well | 46.2 tok/s | 7611 ms | 33K |
| Reasoning | C | Runs well | 46.2 tok/s | 4947 ms | 21K |
How Mixtral 8x7B (47B params) fits at each quantization level on RTX 6000 Ada 48GB (48.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 18.3 GB | Low | D37 |
Q3_K_S | 3 | 23.0 GB | Low | D40 |
NVFP4 | 4 |
ollama run mixtral-8x7bhuggingface-cli download mixtral-8x7b26.3 GB |
| Medium |
| C41 |
Q4_K_M | 4 | 28.7 GB | Medium | C42 |
Q5_K_MBest for your GPU | 5 | 33.8 GB | High | C44 |
Q6_K | 6 | 38.5 GB | High | C44 |
Q8_0 | 8 | 50.3 GB | Very High | F0 |
F16 | 16 | 96.4 GB | Maximum | F0 |