RX 7900 XTX 24GBBudget pick
B59.6 tok/s decode
~$999 MSRP
Can it run?
Tight fit
Using Q4_K_M in llama.cpp
Fit status
Tight fit
Decode
6.7 tok/s
TTFT
28767 ms
Safe context
8K
Memory
18.9 GB / 23.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Tight fit | 7.5 tok/s | 37775 ms | 8K |
| Chat | C | Runs well | 7.5 tok/s | 14166 ms | 8K |
| Coding | C | Tight fit | 6.7 tok/s | 28767 ms | 8K |
| RAG | C | Tight fit | 7.5 tok/s | 47219 ms | 8K |
| Reasoning | C | Tight fit | 7.5 tok/s | 30692 ms | 8K |
How CogVLM2 19B (19B params) fits at each quantization level on MacBook Pro M4 32GB (23.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 7.4 GB | Low | D36 |
Q3_K_S | 3 | 9.3 GB | Low | D38 |
NVFP4 | 4 |
huggingface-cli download cogvlm2-19bUpgrade options
~$999 MSRP
RTX 3090 24GBBest value~$1,499 MSRP
RTX 4090 24GBBiggest leap~$1,599 MSRP
10.6 GB |
| Medium |
| D39 |
Q4_K_M | 4 | 11.6 GB | Medium | D40 |
Q5_K_M | 5 | 13.7 GB | High | C42 |
Q6_KBest for your GPU | 6 | 15.6 GB | High | C44 |
Q8_0 | 8 | 20.3 GB | Very High | C44 |
F16 | 16 | 38.9 GB | Maximum | F0 |
~$2,499 MSRP