RX 7600 8GBBudget pick
C68.5 tok/s decode
~$269 MSRP
Can it run?
Tight fit
Using Q4_K_M in Ollama
Fit status
Tight fit
Decode
85.0 tok/s
TTFT
2278 ms
Safe context
19K
Memory
5.0 GB / 6.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Tight fit | 64.9 tok/s | 4338 ms | 33K |
| Chat | C | Tight fit | 64.9 tok/s | 1627 ms | 10K |
| Coding | C | Tight fit | 85.0 tok/s | 2278 ms | 19K |
| RAG | C | Tight fit | 64.9 tok/s | 5422 ms | 33K |
| Reasoning | C | Tight fit | 64.9 tok/s | 3524 ms | 19K |
How Qwen 3 4B (4B params) fits at each quantization level on GTX 1660 Ti 6GB (6.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 1.6 GB | Low | D34 |
Q3_K_S | 3 | 2.0 GB | Low | D35 |
NVFP4 | 4 |
ollama run qwen-3-4bhuggingface-cli download qwen-3-4bUpgrade options
~$269 MSRP
Intel Arc A750 8GBBest value~$289 MSRP
RTX 2060 Super 8GBNVIDIA upgrade~$399 MSRP
Intel Arc A580 8GBBiggest leap2.2 GB |
| Medium |
| D36 |
Q4_K_M | 4 | 2.4 GB | Medium | D37 |
Q5_K_M | 5 | 2.9 GB | High | D39 |
Q6_KBest for your GPU | 6 | 3.3 GB | High | C40 |
Q8_0 | 8 | 4.3 GB | Very High | C43 |
F16 | 16 | 8.2 GB | Maximum | F0 |