RTX 3060 × Meta Llama 3.1 8B Instruct
Meta Llama 3.1 8B Instruct on RTX 3060
57.5best gen tok/s
1,555best 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 | 57.5 | 1,545 | LocalScore ↗ | 512+128 | 838 ms | llamafile 0.9.2 |
| Q4_K - Medium | 53.1 | 1,487 | LocalScore ↗ | 512+128 | 885 ms | llamafile 0.9.2 |
| Q4_K - Medium | 53.0 | 1,506 | LocalScore ↗ | 512+128 | 864 ms | llamafile 0.9.2 |
| Q4_K - Medium | 52.8 | 1,503 | LocalScore ↗ | 512+128 | 867 ms | llamafile 0.9.2 |
| Q4_K - Medium | 52.4 | 1,478 | LocalScore ↗ | 512+128 | 896 ms | llamafile 0.9.2 |
| Q4_K - Medium | 52.2 | 1,429 | LocalScore ↗ | 512+128 | 907 ms | llamafile 0.9.2 |
| Q4_K - Medium | 52.0 | 1,482 | LocalScore ↗ | 512+128 | 888 ms | llamafile 0.9.2 |
| Q4_K - Medium | 49.7 | 1,400 | LocalScore ↗ | 512+128 | 933 ms | llamafile 0.9.2 |
| Q4_K - Medium | 49.6 | 1,555 | LocalScore ↗ | 512+128 | 845 ms | llamafile 0.9.3 |
| Q4_K - Medium | 49.5 | 1,515 | LocalScore ↗ | 512+128 | 869 ms | llamafile 0.9.2 |
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.
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.