Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
141.6 tok/s
TTFT
1367 ms
Safe context
35K
Memory
5.5 GB / 12.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 141.6 tok/s | 1988 ms | 65K |
| Chat | C | Runs well | 141.6 tok/s | 746 ms | 17K |
| Coding | C | Runs well | 141.6 tok/s | 1367 ms | 35K |
| RAG | C | Runs well | 141.6 tok/s | 2485 ms | 65K |
| Reasoning | C | Runs well | 141.6 tok/s | 1615 ms | 35K |
How Phi 3 Mini 3.8B (3.799999952316284B params) fits at each quantization level on Intel Arc B580 12GB (12.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 1.5 GB | Low | D31 |
Q3_K_S | 3 | 1.9 GB | Low | D32 |
NVFP4 | 4 |
ollama run phi-3-mini-3.8bhuggingface-cli download phi-3-mini-3.8b2.1 GB |
| Medium |
| D32 |
Q4_K_M | 4 | 2.3 GB | Medium | D32 |
Q5_K_M | 5 | 2.7 GB | High | D33 |
Q6_K | 6 | 3.1 GB | High | D34 |
Q8_0 | 8 | 4.1 GB | Very High | D36 |
F16Best for your GPU | 16 | 7.8 GB | Maximum | C42 |