Can it run?
Runs well
Using Q4_K_M in vLLM
Fit status
Runs well
Decode
113.3 tok/s
TTFT
1709 ms
Safe context
22K
Memory
5.8 GB / 8.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | B | Runs well | 92.9 tok/s | 3033 ms | 43K |
| Chat | B | Runs well | 92.9 tok/s | 1137 ms | 11K |
| Coding | B | Runs well | 113.3 tok/s | 1709 ms | 22K |
| RAG | B | Runs well | 92.9 tok/s | 3791 ms | 43K |
| Reasoning | B | Runs well | 92.9 tok/s | 2464 ms | 22K |
How Ministral 3 3B (3B params) fits at each quantization level on Radeon PRO W7600 8GB (8.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 1.2 GB | Low | D32 |
Q3_K_S | 3 | 1.5 GB | Low | D33 |
NVFP4 | 4 |
huggingface-cli download ministral-3-3b| Medium |
| D33 |
Q4_K_M | 4 | 1.8 GB | Medium | D33 |
Q5_K_M | 5 | 2.2 GB | High | D34 |
Q6_K | 6 | 2.5 GB | High | D35 |
Q8_0Best for your GPU | 8 | 3.2 GB | Very High | D37 |
F16 | 16 | 6.1 GB | Maximum | C44 |