MacBook Pro M4 32GBBest value
C5.9 tok/s decode
~$799 MSRP
Can it run?
Too heavy
Using Q4_K_M in Ollama
Fit status
Too heavy
Decode
28.3 tok/s
TTFT
6834 ms
Safe context
12K
Memory
21.2 GB / 16.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | F | Too heavy | 22.8 tok/s | 12340 ms | 21K |
| Chat | F | Too heavy | 22.8 tok/s | 4628 ms | 7K |
| Coding | F | Too heavy | 28.3 tok/s | 6834 ms | 12K |
| RAG | F | Too heavy | 22.8 tok/s | 15425 ms | 21K |
| Reasoning | F | Too heavy | 22.8 tok/s | 10026 ms | 12K |
How Devstral Small 2 24B Instruct (24B params) fits at each quantization level on RX 6950 XT 16GB (16.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_KBest for your GPU | 2 | 9.4 GB | Low | C42 |
Q3_K_S | 3 | 11.8 GB | Low | C44 |
Upgrade options
~$799 MSRP
~$999 MSRP
RTX 5090 32GBBiggest leap| 4 |
13.4 GB |
| Medium |
| C44 |
Q4_K_M | 4 | 14.6 GB | Medium | C44 |
Q5_K_M | 5 | 17.3 GB | High | F0 |
Q6_K | 6 | 19.7 GB | High | F0 |
Q8_0 | 8 | 25.7 GB | Very High | F0 |
F16 | 16 | 49.2 GB | Maximum | F0 |
~$1,999 MSRP