RTX 5070 12GBBudget pickC115.7 tok/s decode
~$549 MSRP
Can it run?
Tight fit
Using Q4_K_M in Ollama
Fit status
Tight fit
Decode
119.6 tok/s
TTFT
1619 ms
Safe context
4K
Memory
6.6 GB / 8.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Tight fit | 119.6 tok/s | 2354 ms | 4K |
| Chat | B | Runs well | 119.6 tok/s | 883 ms | 4K |
| Coding | C | Tight fit | 119.6 tok/s | 1619 ms | 4K |
| RAG | C | Tight fit | 119.6 tok/s | 2943 ms | 4K |
| Reasoning | C | Tight fit | 119.6 tok/s | 1913 ms | 4K |
How Yi 1.5 6B (6B params) fits at each quantization level on RTX 3070 Ti 8GB (8.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 2.3 GB | Low | D35 |
Q3_K_S | 3 | 2.9 GB | Low | D36 |
NVFP4 | 4 | 3.4 GB | Medium | D37 |
Q4_K_M | 4 | 3.7 GB | Medium | D38 |
Q5_K_M | 5 | 4.3 GB | High | D40 |
Q6_KBest for your GPU | 6 | 4.9 GB | High | C42 |
Q8_0 | 8 | 6.4 GB | Very High | C43 |
F16 | 16 | 12.3 GB | Maximum | F0 |
ollama run yi-1.5-6bhuggingface-cli download yi-1.5-6bUpgrade options
RTX 5070 12GBBudget pick~$549 MSRP
RTX 4070 12GBBest value~$599 MSRP
RTX 3080 10GBBiggest leap~$699 MSRP
RTX 2080 Ti 11GBNVIDIA upgrade~$999 MSRP