RX 9070 16GBBudget pick
C81.3 tok/s decode
~$479 MSRP
Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
20.0 tok/s
TTFT
9690 ms
Safe context
22K
Memory
9.3 GB / 13.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 22.4 tok/s | 12550 ms | 39K |
| Chat | C | Runs well | 22.4 tok/s | 4706 ms | 12K |
| Coding | C | Runs well | 20.0 tok/s | 9690 ms | 22K |
| RAG | C | Runs well | 22.4 tok/s | 15687 ms | 39K |
| Reasoning | C | Runs well | 22.4 tok/s | 10197 ms | 22K |
How DeepSeek R1 0528 Qwen3 8B (8B params) fits at each quantization level on MacBook Pro M3 Pro 18GB (13.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 3.1 GB | Low | D35 |
Q3_K_S | 3 | 3.9 GB | Low | D36 |
NVFP4 | 4 |
Upgrade options
~$479 MSRP
~$499 MSRP
4.5 GB |
| Medium |
| D37 |
Q4_K_M | 4 | 4.9 GB | Medium | D38 |
Q5_K_M | 5 | 5.8 GB | High | D39 |
Q6_K | 6 | 6.6 GB | High | C40 |
Q8_0Best for your GPU | 8 | 8.6 GB | Very High | C44 |
F16 | 16 | 16.4 GB | Maximum | F0 |
~$799 MSRP