NVIDIA A100 40GBBudget pickC89.2 tok/s decode
~$10,000 MSRP
Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
12.1 tok/s
TTFT
15972 ms
Safe context
22K
Memory
24.8 GB / 34.6 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Tight fit | 14.4 tok/s | 19612 ms | 39K |
| Chat | C | Runs well | 14.4 tok/s | 7354 ms | 12K |
| Coding | C | Runs well | 12.1 tok/s | 15972 ms | 22K |
| RAG | C | Tight fit | 14.4 tok/s | 24515 ms | 39K |
| Reasoning | C | Runs well | 14.4 tok/s | 15935 ms | 22K |
How Devstral Small 2 24B Instruct (24B params) fits at each quantization level on MacBook Pro M4 Pro 48GB (34.6 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 9.4 GB | Low | D35 |
Q3_K_S | 3 | 11.8 GB | Low | D37 |
NVFP4 | 4 |
ollama run devstral-small-2-24bhuggingface-cli download devstral-small-2-24bUpgrade options
NVIDIA A100 40GBBudget pick~$10,000 MSRP
RTX PRO 5000 Blackwell 48GBBiggest leap
13.4 GB |
| Medium |
| D38 |
Q4_K_M | 4 | 14.6 GB | Medium | D39 |
Q5_K_M | 5 | 17.3 GB | High | C40 |
Q6_K | 6 | 19.7 GB | High | C42 |
Q8_0Best for your GPU | 8 | 25.7 GB | Very High | C44 |
F16 | 16 | 49.2 GB | Maximum | F0 |