Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
54.0 tok/s
TTFT
3588 ms
Safe context
16K
Memory
13.9 GB / 20.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 54.0 tok/s | 5219 ms | 16K |
| Chat | C | Runs well | 54.0 tok/s | 1957 ms | 12K |
| Coding | C | Runs well | 54.0 tok/s | 3588 ms | 16K |
| RAG | C | Runs well | 54.0 tok/s | 6524 ms | 16K |
| Reasoning | C | Runs well | 54.0 tok/s | 4240 ms | 16K |
How Phi-4 14B (14B params) fits at each quantization level on RTX A4500 20GB (20.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 5.5 GB | Low | D35 |
Q3_K_S | 3 | 6.9 GB | Low | D36 |
NVFP4 | 4 |
ollama run phi-4-14bhuggingface-cli download phi-4-14b| Medium |
| D37 |
Q4_K_M | 4 | 8.5 GB | Medium | D38 |
Q5_K_M | 5 | 10.1 GB | High | D40 |
Q6_K | 6 | 11.5 GB | High | C41 |
Q8_0Best for your GPU | 8 | 15.0 GB | Very High | C44 |
F16 | 16 | 28.7 GB | Maximum | F0 |