Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
83.4 tok/s
TTFT
2321 ms
Safe context
29K
Memory
22.1 GB / 40.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | B | Runs well | 97.3 tok/s | 2893 ms | 50K |
| Chat | C | Runs well | 97.3 tok/s | 1085 ms | 16K |
| Coding | C | Runs well | 83.4 tok/s | 2321 ms | 29K |
| RAG | B | Runs well | 97.3 tok/s | 3616 ms | 50K |
| Reasoning | C | Runs well | 97.3 tok/s | 2351 ms | 29K |
How Codestral 22B v0.1 IMat (22B params) fits at each quantization level on NVIDIA A100 40GB (40.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 8.6 GB | Low | D33 |
Q3_K_S | 3 | 10.8 GB | Low | D35 |
NVFP4 | 4 |
| Medium |
| D35 |
Q4_K_M | 4 | 13.4 GB | Medium | D36 |
Q5_K_M | 5 | 15.8 GB | High | D37 |
Q6_K | 6 | 18.0 GB | High | D38 |
Q8_0Best for your GPU | 8 | 23.5 GB | Very High | C41 |
F16 | 16 | 45.1 GB | Maximum | F0 |
