Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
100.7 tok/s
TTFT
1922 ms
Safe context
25K
Memory
20.5 GB / 32.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | B | Runs well | 100.7 tok/s | 2796 ms | 43K |
| Chat | C | Runs well | 100.7 tok/s | 1048 ms | 14K |
| Coding | B | Runs well | 100.7 tok/s | 1922 ms | 25K |
| RAG | B | Runs well | 100.7 tok/s | 3495 ms | 43K |
| Reasoning | B | Runs well | 100.7 tok/s | 2272 ms | 25K |
How Codestral 21B Pruned i1 (21B params) fits at each quantization level on RTX 5090 32GB (32.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 8.2 GB | Low | D34 |
Q3_K_S | 3 | 10.3 GB | Low | D36 |
NVFP4 | 4 |
11.8 GB |
| Medium |
| D37 |
Q4_K_M | 4 | 12.8 GB | Medium | D37 |
Q5_K_M | 5 | 15.1 GB | High | D39 |
Q6_K | 6 | 17.2 GB | High | C40 |
Q8_0Best for your GPU | 8 | 22.5 GB | Very High | C44 |
F16 | 16 | 43.1 GB | Maximum | F0 |