Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
50.1 tok/s
TTFT
3861 ms
Safe context
28K
Memory
46.0 GB / 80.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 58.5 tok/s | 4814 ms | 48K |
| Chat | C | Runs well | 58.5 tok/s | 1805 ms | 15K |
| Coding | C | Runs well | 50.1 tok/s | 3861 ms | 28K |
| RAG | C | Runs well | 58.5 tok/s | 6018 ms | 48K |
| Reasoning | C | Runs well | 58.5 tok/s | 3911 ms | 28K |
How Qwen3 48B A4B Savant Commander Distill 12X Closed Open Heretic Uncensored (48B params) fits at each quantization level on NVIDIA A100 80GB (80.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 18.7 GB | Low | D34 |
Q3_K_S | 3 | 23.5 GB | Low | D35 |
NVFP4 | 4 |
26.9 GB |
| Medium |
| D36 |
Q4_K_M | 4 | 29.3 GB | Medium | D37 |
Q5_K_M | 5 | 34.6 GB | High | D38 |
Q6_K | 6 | 39.4 GB | High | D40 |
Q8_0Best for your GPU | 8 | 51.4 GB | Very High | C43 |
F16 | 16 | 98.4 GB | Maximum | F0 |