Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
153.8 tok/s
TTFT
1259 ms
Safe context
25K
Memory
7.8 GB / 12.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | B | Runs well | 158.1 tok/s | 1782 ms | 43K |
| Chat | B | Runs well | 158.1 tok/s | 668 ms | 13K |
| Coding | B | Runs well | 153.8 tok/s | 1259 ms | 25K |
| RAG | B | Runs well | 158.1 tok/s | 2227 ms | 43K |
| Reasoning | B | Runs well | 158.1 tok/s | 1448 ms | 25K |
How falcon mamba 7b instruct Q4 K M (7B params) fits at each quantization level on RTX 3080 Ti 12GB (12.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 2.7 GB | Low | D34 |
Q3_K_S | 3 | 3.4 GB | Low | D35 |
NVFP4 | 4 |
3.9 GB |
| Medium |
| D36 |
Q4_K_M | 4 | 4.3 GB | Medium | D37 |
Q5_K_M | 5 | 5.0 GB | High | D38 |
Q6_K | 6 | 5.7 GB | High | D39 |
Q8_0Best for your GPU | 8 | 7.5 GB | Very High | C42 |
F16 | 16 | 14.3 GB | Maximum | F0 |
