Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
276.6 tok/s
TTFT
700 ms
Safe context
34K
Memory
5.6 GB / 12.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 276.6 tok/s | 1018 ms | 63K |
| Chat | C | Runs well | 276.6 tok/s | 382 ms | 17K |
| Coding | C | Runs well | 276.6 tok/s | 700 ms | 34K |
| RAG | C | Runs well | 276.6 tok/s | 1273 ms | 63K |
| Reasoning | C | Runs well | 276.6 tok/s | 827 ms | 34K |
How gemma 3 4b it (4B params) fits at each quantization level on RTX 3080 Ti 12GB (12.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 1.6 GB | Low | D32 |
Q3_K_S | 3 | 2.0 GB | Low | D33 |
NVFP4 | 4 | 2.2 GB | Medium | D33 |
Q4_K_M | 4 | 2.4 GB | Medium | D34 |
Q5_K_M | 5 | 2.9 GB | High | D35 |
Q6_K | 6 | 3.3 GB | High | D35 |
Q8_0 | 8 | 4.3 GB | Very High | D37 |
F16Best for your GPU | 16 | 8.2 GB | Maximum | C44 |
