AMD Instinct MI350X 288GBBudget pick
B108.7 tok/s decode
~$8,000 MSRP
Can it run?
Tight fit
Using Q4_K_M in llama.cpp
Fit status
Tight fit
Decode
125.0 tok/s
TTFT
1548 ms
Safe context
18K
Memory
166.9 GB / 192.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Tight fit | 125.0 tok/s | 2252 ms | 36K |
| Chat | C | Tight fit | 125.0 tok/s | 845 ms | 9K |
| Coding | C | Tight fit | 125.0 tok/s | 1548 ms | 18K |
| RAG | C | Tight fit | 125.0 tok/s | 2815 ms | 36K |
| Reasoning | C | Tight fit | 125.0 tok/s | 1830 ms | 18K |
How Qwen 3 235B A22B (235B params) fits at each quantization level on NVIDIA GB200 192GB (192.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 91.7 GB | Low | D40 |
Q3_K_S | 3 | 115.2 GB | Low | C42 |
NVFP4 | 4 |
huggingface-cli download qwen-3-235b-a22bUpgrade options
~$8,000 MSRP
131.6 GB |
| Medium |
| C44 |
Q4_K_MBest for your GPU | 4 | 143.4 GB | Medium | C45 |
Q5_K_M | 5 | 169.2 GB | High | C45 |
Q6_K | 6 | 192.7 GB | High | F0 |
Q8_0 | 8 | 251.5 GB | Very High | F0 |
F16 | 16 | 481.7 GB | Maximum | F0 |