MacBook Pro M4 32GBBest value
C6.4 tok/s decode
~$799 MSRP
Can it run?
Too heavy
Using Q4_K_M in vLLM
Fit status
Too heavy
Decode
21.7 tok/s
TTFT
8903 ms
Safe context
12K
Memory
20.9 GB / 16.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | F | Too heavy | 21.7 tok/s | 12949 ms | 21K |
| Chat | D | Very compromised (needs ~2.2 GB host RAM) | 18.7 tok/s | 5642 ms | 7K |
| Coding | F | Too heavy | 21.7 tok/s | 8903 ms | 12K |
| RAG | F | Too heavy | 21.7 tok/s | 16186 ms | 21K |
| Reasoning | F | Too heavy | 21.7 tok/s | 10521 ms | 12K |
How Codestral 2 25.08 (22B params) fits at each quantization level on RX 6900 XT 16GB (16.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 8.6 GB | Low | C41 |
Q3_K_SBest for your GPU | 3 | 10.8 GB | Low | C44 |
NVFP4 | 4 | 12.3 GB | Medium | C44 |
Q4_K_M | 4 | 13.4 GB | Medium | C44 |
Q5_K_M | 5 | 15.8 GB | High | C44 |
Q6_K | 6 | 18.0 GB | High | F0 |
Q8_0 | 8 | 23.5 GB | Very High | F0 |
F16 | 16 | 45.1 GB | Maximum | F0 |
Upgrade options
~$799 MSRP
~$999 MSRP
RTX 5090 32GBBiggest leap~$1,999 MSRP