AMD Instinct MI350X 288GBBudget pick
C90.9 tok/s decode
~$8,000 MSRP
Can it run?
Runs well
Using Q3_K_M in Ollama
Fit status
Runs well
Decode
68.1 tok/s
TTFT
2841 ms
Safe context
39K
Memory
105.6 GB / 256.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 68.1 tok/s | 4133 ms | 66K |
| Chat | C | Runs well | 68.1 tok/s | 1550 ms | 21K |
| Coding | C | Runs well | 68.1 tok/s | 2841 ms | 39K |
| RAG | C | Runs well | 68.1 tok/s | 5166 ms | 66K |
| Reasoning | C | Runs well | 67.2 tok/s | 3489 ms | 39K |
How Qwen3.5 122B A10B (122B params) fits at each quantization level on AMD Instinct MI325X 256GB (256.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 47.6 GB | Low | D33 |
Q3_K_S | 3 | 59.8 GB | Low | D35 |
NVFP4 | 4 |
huggingface-cli download hf-unsloth--qwen3-5-122b-a10b-ggufUpgrade options
~$8,000 MSRP
68.3 GB |
| Medium |
| D35 |
Q4_K_M | 4 | 74.4 GB | Medium | D36 |
Q5_K_M | 5 | 87.8 GB | High | D37 |
Q6_K | 6 | 100.0 GB | High | D38 |
Q8_0Best for your GPU | 8 | 130.5 GB | Very High | C40 |
F16 | 16 | 250.1 GB | Maximum | C45 |