AMD Instinct MI350X 288GBBudget pick
C398.9 tok/s decode
~$8,000 MSRP
Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
36.4 tok/s
TTFT
5324 ms
Safe context
62K
Memory
47.2 GB / 184.3 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 36.4 tok/s | 7744 ms | 116K |
| Chat | C | Runs well | 36.4 tok/s | 2904 ms | 33K |
| Coding | C | Runs well | 36.4 tok/s | 5324 ms | 62K |
| RAG | C | Runs well | 36.4 tok/s | 9680 ms | 116K |
| Reasoning | C | Runs well | 36.4 tok/s | 6292 ms | 62K |
How Devstral Small 2 24B Instruct (24B params) fits at each quantization level on Mac Studio M3 Ultra 256GB (184.3 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 9.4 GB | Low | D31 |
Q3_K_S | 3 | 11.8 GB | Low | D31 |
NVFP4 | 4 |
ollama run devstral-small-2-24bhuggingface-cli download devstral-small-2-24bUpgrade options
~$8,000 MSRP
~$15,000 MSRP
H100 NVL 188GBBiggest leap13.4 GB |
| Medium |
| D31 |
Q4_K_M | 4 | 14.6 GB | Medium | D31 |
Q5_K_M | 5 | 17.3 GB | High | D31 |
Q6_K | 6 | 19.7 GB | High | D32 |
Q8_0 | 8 | 25.7 GB | Very High | D32 |
F16Best for your GPU | 16 | 49.2 GB | Maximum | D35 |