Can it run?
Too heavy
Using Q4_K_M in vLLM
Fit status
Too heavy
Decode
4.9 tok/s
TTFT
39451 ms
Safe context
4K
Memory
422.3 GB / 48.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | F | Too heavy | 5.6 tok/s | 50385 ms | 4K |
| Chat | F | Too heavy | 5.6 tok/s | 18894 ms | 4K |
| Coding | F | Too heavy | 4.9 tok/s | 39451 ms | 4K |
| RAG | F | Too heavy | 5.6 tok/s | 62981 ms | 4K |
| Reasoning | F | Too heavy | 5.6 tok/s | 40938 ms | 4K |
How DeepSeek V3 671B (671B params) fits at each quantization level on RTX 6000 Ada 48GB (48.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 261.7 GB | Low | F0 |
Q3_K_S | 3 | 328.8 GB | Low | F0 |
NVFP4 | 4 |
| Medium |
| F0 |
Q4_K_M | 4 | 409.3 GB | Medium | F0 |
Q5_K_M | 5 | 483.1 GB | High | F0 |
Q6_K | 6 | 550.2 GB | High | F0 |
Q8_0 | 8 | 718.0 GB | Very High | F0 |
F16 | 16 | 1375.6 GB | Maximum | F0 |
