RX 7900 XTX 24GBBudget pick
C326.4 tok/s decode
~$999 MSRP
Can it run?
Runs well
Using Q5_K_M in Ollama
Fit status
Runs well
Decode
40.8 tok/s
TTFT
4745 ms
Safe context
48K
Memory
7.6 GB / 23.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 40.8 tok/s | 6902 ms | 95K |
| Chat | C | Runs well | 40.8 tok/s | 2588 ms | 24K |
| Coding | C | Runs well | 40.8 tok/s | 4745 ms | 48K |
| RAG | C | Runs well | 40.8 tok/s | 8627 ms | 95K |
| Reasoning | C | Runs well | 40.8 tok/s | 5608 ms | 48K |
How Llama 3.2 3B Instruct (3B params) fits at each quantization level on MacBook Pro M4 32GB (23.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 1.2 GB | Low | D31 |
Q3_K_S | 3 | 1.5 GB | Low | D31 |
NVFP4 | 4 | 1.7 GB | Medium | D31 |
Q4_K_M | 4 | 1.8 GB | Medium | D31 |
Q5_K_M | 5 | 2.2 GB | High | D32 |
Q6_K | 6 | 2.5 GB | High | D32 |
Q8_0 | 8 | 3.2 GB | Very High | D32 |
F16Best for your GPU | 16 | 6.1 GB | Maximum | D35 |
huggingface-cli download hf-bartowski--llama-3-2-3b-instruct-ggufUpgrade options
~$999 MSRP
RTX 3090 24GBBest value~$1,499 MSRP
~$2,499 MSRP
~$2,499 MSRP