RX 7900 XT 20GBBudget pick
C98.4 tok/s decode
~$899 MSRP
Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
13.9 tok/s
TTFT
13894 ms
Safe context
28K
Memory
9.9 GB / 17.3 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 13.9 tok/s | 20209 ms | 49K |
| Chat | C | Runs well | 13.9 tok/s | 7578 ms | 15K |
| Coding | C | Runs well | 13.9 tok/s | 13894 ms | 28K |
| RAG | C | Runs well | 13.9 tok/s | 25261 ms | 49K |
| Reasoning | C | Runs well | 13.9 tok/s | 16420 ms | 28K |
How Meta Llama 3 8B Instruct (8B params) fits at each quantization level on MacBook Air M3 24GB (17.3 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 3.1 GB | Low | D33 |
Q3_K_S | 3 | 3.9 GB | Low | D34 |
NVFP4 | 4 | 4.5 GB | Medium | D35 |
Q4_K_M | 4 | 4.9 GB | Medium | D35 |
Q5_K_M | 5 | 5.8 GB | High | D37 |
Q6_K | 6 | 6.6 GB | High | D38 |
Q8_0Best for your GPU | 8 | 8.6 GB | Very High | C40 |
F16 | 16 | 16.4 GB | Maximum | C44 |
Upgrade options
~$899 MSRP
~$999 MSRP
~$1,999 MSRP
RTX A4500 20GBBiggest leap