AMD Instinct MI350X 288GBBudget pick
C273.5 tok/s decode
~$8,000 MSRP
Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
16.1 tok/s
TTFT
12016 ms
Safe context
35K
Memory
41.8 GB / 92.2 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 16.1 tok/s | 17478 ms | 62K |
| Chat | C | Runs well | 16.1 tok/s | 6554 ms | 19K |
| Coding | C | Runs well | 16.1 tok/s | 12016 ms | 35K |
| RAG | C | Runs well | 16.1 tok/s | 21848 ms | 62K |
| Reasoning | C | Runs well | 16.1 tok/s | 14201 ms | 35K |
How Qwen3.5 35B A3B (35B params) fits at each quantization level on MacBook Pro M4 Max 128GB (92.2 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 13.7 GB | Low | D33 |
Q3_K_S | 3 | 17.2 GB | Low | D34 |
NVFP4 | 4 | 19.6 GB | Medium | D34 |
Q4_K_M | 4 | 21.3 GB | Medium | D35 |
Q5_K_M | 5 | 25.2 GB | High | D35 |
Q6_K | 6 | 28.7 GB | High | D36 |
Q8_0 | 8 | 37.5 GB | Very High | D38 |
F16Best for your GPU | 16 | 71.8 GB | Maximum | C45 |
huggingface-cli download hf-unsloth--qwen3-5-35b-a3b-ggufUpgrade options
~$8,000 MSRP
~$15,000 MSRP
NVIDIA GH200 96GBBiggest leap