Can it run?
Runs well
Using Q4_K_M in vLLM
Fit status
Runs well
Decode
109.3 tok/s
TTFT
1771 ms
Safe context
26K
Memory
178.4 GB / 288.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | B | Runs well | 109.3 tok/s | 2577 ms | 51K |
| Chat | B | Runs well | 109.3 tok/s | 966 ms | 13K |
| Coding | B | Runs well | 109.3 tok/s | 1771 ms | 26K |
| RAG | B | Runs well | 109.3 tok/s | 3221 ms | 51K |
| Reasoning | B | Runs well | 109.3 tok/s | 2094 ms | 26K |
How DeepSeek Coder V2 236B (236B params) fits at each quantization level on AMD Instinct MI350X 288GB (288.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 92.0 GB | Low | D36 |
Q3_K_S | 3 | 115.6 GB | Low | D38 |
NVFP4 | 4 | 132.2 GB | Medium | D39 |
Q4_K_M | 4 | 144.0 GB | Medium | C40 |
Q5_K_M | 5 | 169.9 GB | High | C42 |
Q6_KBest for your GPU | 6 | 193.5 GB | High | C44 |
Q8_0 | 8 | 252.5 GB | Very High | C44 |
F16 | 16 | 483.8 GB | Maximum | F0 |
huggingface-cli download deepseek-coder-v2-236b