RX 7900 XTX 24GBBudget pick
C141.6 tok/s decode
~$999 MSRP
Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
47.5 tok/s
TTFT
4072 ms
Safe context
34K
Memory
10.8 GB / 23.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 47.5 tok/s | 5923 ms | 61K |
| Chat | C | Runs well | 47.5 tok/s | 2221 ms | 18K |
| Coding | C | Runs well | 47.5 tok/s | 4072 ms | 34K |
| RAG | C | Runs well | 47.5 tok/s | 7404 ms | 61K |
| Reasoning | C | Runs well | 47.5 tok/s | 4813 ms | 34K |
How Meta Llama 3.1 8B Instruct (8B params) fits at each quantization level on MacBook Pro M2 Max 32GB (23.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 3.1 GB | Low | D32 |
Q3_K_S | 3 | 3.9 GB | Low | D33 |
NVFP4 | 4 | 4.5 GB | Medium | D34 |
Q4_K_M | 4 | 4.9 GB | Medium | D34 |
Q5_K_M | 5 | 5.8 GB | High | D35 |
Q6_K | 6 | 6.6 GB | High | D36 |
Q8_0 | 8 | 8.6 GB | Very High | D38 |
F16Best for your GPU | 16 | 16.4 GB | Maximum | C45 |
Upgrade options
~$999 MSRP
RTX 3090 24GBBest value~$1,499 MSRP
NVIDIA A10 24GBBiggest leap