RTX 3060 Ti × Meta Llama 3.1 8B Instruct
Meta Llama 3.1 8B Instruct on RTX 3060 Ti
67.3best gen tok/s
2,084best prompt tok/s
640max context tested
10results
Measured · all numbers below come from the linked sources
llama.cpp
| Quant | Gen tok/s | Prompt tok/s | Source | Context tested | TTFT | Build |
|---|---|---|---|---|---|---|
| Q4_K - Medium | 67.3 | 1,974 | LocalScore ↗ | 512+128 | 651 ms | llamafile 0.9.2 |
| Q4_K - Medium | 65.3 | 1,870 | LocalScore ↗ | 512+128 | 689 ms | llamafile 0.9.3 |
| Q4_K - Medium | 65.1 | 1,898 | LocalScore ↗ | 512+128 | 683 ms | llamafile 0.9.2 |
| Q4_K - Medium | 63.8 | 2,084 | LocalScore ↗ | 512+128 | 624 ms | llamafile 0.9.3 |
| Q4_K - Medium | 63.4 | 1,830 | LocalScore ↗ | 512+128 | 712 ms | llamafile 0.9.2 |
| Q4_K - Medium | 61.7 | 1,729 | LocalScore ↗ | 512+128 | 767 ms | llamafile 0.9.2 |
| Q4_K - Medium | 59.2 | 1,905 | LocalScore ↗ | 512+128 | 678 ms | llamafile 0.9.3 |
| Q4_K - Medium | 58.6 | 1,962 | LocalScore ↗ | 512+128 | 657 ms | llamafile 0.9.2 |
| Q4_K - Medium | 54.4 | 1,861 | LocalScore ↗ | 512+128 | 696 ms | llamafile 0.9.2 |
| Q4_K - Medium | 42.6 | 1,279 | LocalScore ↗ | 512+128 | 1,079 ms | llamafile 0.9.3 |
Ollama
No measured Ollama results yet. Submit yours →
MLX
No measured MLX results yet. Submit yours →
"Context tested" is the largest prompt+generation length actually benchmarked (tokens), not a memory-fit estimate.
What else can RTX 3060 Ti run? →
Includes results from LocalScore, a Mozilla Builders project by cjpais, mirrored in the cn0303/local-llm-speed-benchmarks dataset. LocalScore runs use llamafile, which is built on llama.cpp.