Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
77.4 tok/s
TTFT
2502 ms
Safe context
24K
Memory
5.2 GB / 8.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | B | Runs well | 77.4 tok/s | 3639 ms | 45K |
| Chat | B | Runs well | 77.4 tok/s | 1365 ms | 12K |
| Coding | B | Runs well | 77.4 tok/s | 2502 ms | 24K |
| RAG | B | Runs well | 77.4 tok/s | 4549 ms | 45K |
| Reasoning | B | Runs well | 77.4 tok/s | 2957 ms | 24K |
How gemma 3 4b it (4B params) fits at each quantization level on GTX 1080 8GB (8.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 1.6 GB | Low | D34 |
Q3_K_S | 3 | 2.0 GB | Low | D35 |
NVFP4 | 4 | 2.2 GB | Medium | D35 |
Q4_K_M | 4 | 2.4 GB | Medium | D36 |
Q5_K_M | 5 | 2.9 GB | High | D37 |
Q6_K | 6 | 3.3 GB | High | D38 |
Q8_0Best for your GPU | 8 | 4.3 GB | Very High | C41 |
F16 | 16 | 8.2 GB | Maximum | F0 |
