Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
131.1 tok/s
TTFT
1742 ms
Safe context
54K
Memory
85.2 GB / 288.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 131.1 tok/s | 2534 ms | 96K |
| Chat | C | Runs well | 131.1 tok/s | 950 ms | 29K |
| Coding | C | Runs well | 131.1 tok/s | 1742 ms | 54K |
| RAG | C | Runs well | 131.1 tok/s | 3168 ms | 96K |
| Reasoning | C | Runs well | 131.1 tok/s | 2059 ms | 54K |
How Qwen 2.5 72B (72B params) fits at each quantization level on AMD Instinct MI350X 288GB (288.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 28.1 GB | Low | D32 |
Q3_K_S | 3 | 35.3 GB | Low | D32 |
NVFP4 | 4 |
ollama run qwen-2.5-72bhuggingface-cli download qwen-2.5-72b40.3 GB |
| Medium |
| D33 |
Q4_K_M | 4 | 43.9 GB | Medium | D33 |
Q5_K_M | 5 | 51.8 GB | High | D33 |
Q6_K | 6 | 59.0 GB | High | D34 |
Q8_0 | 8 | 77.0 GB | Very High | D35 |
F16Best for your GPU | 16 | 147.6 GB | Maximum | C40 |