AMD Instinct MI350X 288GBBudget pick
B108.7 tok/s decode
~$8,000 MSRP
Can it run?
Too heavy
Using Q4_K_M in llama.cpp
Fit status
Too heavy
Decode
21.4 tok/s
TTFT
9048 ms
Safe context
10K
Memory
157.3 GB / 96.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | F | Too heavy | 25.0 tok/s | 11281 ms | 19K |
| Chat | F | Too heavy | 25.0 tok/s | 4230 ms | 5K |
| Coding | F | Too heavy | 21.4 tok/s | 9048 ms | 10K |
| RAG | F | Too heavy | 25.0 tok/s | 14101 ms | 19K |
| Reasoning | F | Too heavy | 25.0 tok/s | 9166 ms | 10K |
How Qwen 3 235B A22B (235B params) fits at each quantization level on RTX PRO 6000 Blackwell Server Edition 96GB (96.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 91.7 GB | Low | C45 |
Q3_K_S | 3 | 115.2 GB | Low | F0 |
NVFP4 | 4 |
Upgrade options
~$8,000 MSRP
~$15,000 MSRP
131.6 GB |
| Medium |
| F0 |
Q4_K_M | 4 | 143.4 GB | Medium | F0 |
Q5_K_M | 5 | 169.2 GB | High | F0 |
Q6_K | 6 | 192.7 GB | High | F0 |
Q8_0 | 8 | 251.5 GB | Very High | F0 |
F16 | 16 | 481.7 GB | Maximum | F0 |
B100 192GBNVIDIA upgrade