Can it run?
Runs well
Using Q4_K_M in llama.cpp
Fit status
Runs well
Decode
80.4 tok/s
TTFT
2408 ms
Safe context
24K
Memory
173.3 GB / 256.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | B | Runs well | 81.5 tok/s | 3455 ms | 46K |
| Chat | B | Runs well | 81.5 tok/s | 1296 ms | 12K |
| Coding | B | Runs well | 80.4 tok/s | 2408 ms | 24K |
| RAG | B | Runs well | 81.5 tok/s | 4319 ms | 46K |
| Reasoning | B | Runs well | 80.4 tok/s | 2846 ms | 24K |
How Qwen 3 235B A22B (235B params) fits at each quantization level on AMD Instinct MI325X 256GB (256.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 91.7 GB | Low | D37 |
Q3_K_S | 3 | 115.2 GB | Low | D39 |
NVFP4 | 4 |
huggingface-cli download qwen-3-235b-a22b131.6 GB |
| Medium |
| C41 |
Q4_K_M | 4 | 143.4 GB | Medium | C42 |
Q5_K_M | 5 | 169.2 GB | High | C44 |
Q6_KBest for your GPU | 6 | 192.7 GB | High | C45 |
Q8_0 | 8 | 251.5 GB | Very High | C45 |
F16 | 16 | 481.7 GB | Maximum | F0 |