RTX 3060 vs RTX 5070
Best sourced output tok/s per model (single-stream generation) — engine and version, quantization, context and a source link on every row. Hover a value for the quant.
RTX 306013 sourced runs · best 310.4 tok/s
RTX 50705 verified runs · best 185.8 tok/s
Specs and evidence
| RTX 3060 | RTX 5070 | |
|---|---|---|
| VRAM | 12 GB | 12 GB |
| Sourced runs on the board | 13 | 5 |
| Models measured | 12 | 5 |
| Best output tok/s | 310.4 tok/s | 185.8 tok/s |
| Highest measured peak VRAM | 9.3 GB | 9.0 GB |
| Street price (latest snapshot) | — | — |
The measured matrix
| Model | RTX 3060 | RTX 5070 |
|---|---|---|
| Qwen3-0.6B | 310.4 | — |
| Spark-X2.5-1.7B | 207.3 | — |
| NVIDIA-Nemotron-3-Nano-4B | — | 185.8 |
| Apodex-1.1-mini | — | 182.3 |
| gemma-4-E2B-it | 113.4 | — |
| Ornith-1.5-9B | — | 102.0 |
| gemma-4-12B-it | 89.7 | — |
| Spark-X2.5-4B | 84.7 | — |
| Qwen3.8-Flash-Next | — | 81.8 |
| Qwen3.5-4B | 78.6 | — |
| gemma-4-E4B-it | 70.9 | — |
| Ling-3.0-tiny | 52.6 | — |
| granite-4.2-8b | 52.5 | — |
| Ornith-1.0-9B | 52.5 | — |
| Qwen3.5-9B | 51.9 | — |
| Qwythos-9B-v2 | 51.1 | — |
| Qwen3.8-27B | — | 48.3 |