Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
60.7 tok/s
TTFT
3189 ms
Safe context
37K
Memory
8.6 GB / 20.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 65.8 tok/s | 4282 ms | 66K |
| Chat | C | Runs well | 65.8 tok/s | 1606 ms | 19K |
| Coding | C | Runs well | 60.7 tok/s | 3189 ms | 37K |
| RAG | C | Runs well | 65.8 tok/s | 5353 ms | 66K |
| Reasoning | C | Runs well | 65.8 tok/s | 3479 ms | 37K |
How Mistral 7B Instruct v0.2 (7B params) fits at each quantization level on RTX 4000 Ada 20GB (20.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 2.7 GB | Low | D32 |
Q3_K_S | 3 | 3.4 GB | Low | D33 |
NVFP4 | 4 |
| Medium |
| D34 |
Q4_K_M | 4 | 4.3 GB | Medium | D34 |
Q5_K_M | 5 | 5.0 GB | High | D35 |
Q6_K | 6 | 5.7 GB | High | D36 |
Q8_0 | 8 | 7.5 GB | Very High | D37 |
F16Best for your GPU | 16 | 14.3 GB | Maximum | C44 |
