Can it run?
Runs well
Using Q4_K_M in vLLM
Fit status
Runs well
Decode
89.6 tok/s
TTFT
2162 ms
Safe context
25K
Memory
114.6 GB / 180.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | B | Runs well | 89.6 tok/s | 3144 ms | 43K |
| Chat | C | Runs well | 89.6 tok/s | 1179 ms | 14K |
| Coding | B | Runs well | 89.6 tok/s | 2162 ms | 25K |
| RAG | B | Runs well | 89.6 tok/s | 3930 ms | 43K |
| Reasoning | B | Runs well | 89.6 tok/s | 2555 ms | 25K |
How Devstral 2 123B Instruct (123B params) fits at each quantization level on NVIDIA B200 180GB (180.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 48.0 GB | Low | D35 |
Q3_K_S | 3 | 60.3 GB | Low | D37 |
NVFP4 | 4 | 68.9 GB | Medium | D38 |
Q4_K_M | 4 | 75.0 GB | Medium | D38 |
Q5_K_M | 5 | 88.6 GB | High | C40 |
Q6_K | 6 | 100.9 GB | High | C42 |
Q8_0Best for your GPU | 8 | 131.6 GB | Very High | C45 |
F16 | 16 | 252.2 GB | Maximum | F0 |
huggingface-cli download devstral-2-123b