Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
192.2 tok/s
TTFT
1007 ms
Safe context
46K
Memory
27.6 GB / 80.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 192.2 tok/s | 1465 ms | 82K |
| Chat | C | Runs well | 192.2 tok/s | 549 ms | 25K |
| Coding | C | Runs well | 192.2 tok/s | 1007 ms | 46K |
| RAG | C | Runs well | 192.2 tok/s | 1831 ms | 82K |
| Reasoning | C | Runs well | 192.2 tok/s | 1190 ms | 46K |
How Devstral Small 2 24B Instruct (24B params) fits at each quantization level on NVIDIA H100 80GB (80.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 9.4 GB | Low | D32 |
Q3_K_S | 3 | 11.8 GB | Low | D33 |
NVFP4 | 4 | 13.4 GB | Medium | D33 |
Q4_K_M | 4 | 14.6 GB | Medium | D33 |
Q5_K_M | 5 | 17.3 GB | High | D34 |
Q6_K | 6 | 19.7 GB | High | D35 |
Q8_0 | 8 | 25.7 GB | Very High | D36 |
F16Best for your GPU | 16 | 49.2 GB | Maximum | C43 |
ollama run devstral-small-2-24bhuggingface-cli download devstral-small-2-24b