RTX 3080 × Llama 3.2 1B Instruct
Llama 3.2 1B Instruct on RTX 3080
330best gen tok/s
12,562best prompt tok/s
4,352max context tested
15results
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 | 330 | 12,190 | LocalScore ↗ | 4,352 | 108 ms | llamafile 0.9.2 |
| Q4_K - Medium | 322 | 12,425 | LocalScore ↗ | 4,352 | 109 ms | llamafile 0.9.2 |
| Q4_K - Medium | 315 | 12,562 | LocalScore ↗ | 512+128 | 107 ms | llamafile 0.9.2 |
| Q4_K - Medium | 308 | 11,982 | LocalScore ↗ | 512+128 | 111 ms | llamafile 0.9.2 |
| Q4_K - Medium | 235 | 11,682 | LocalScore ↗ | 512+128 | 113 ms | llamafile 0.9.3 |
| Q4_K - Medium | 233 | 11,173 | LocalScore ↗ | 512+128 | 117 ms | llamafile 0.9.3 |
| Q4_K - Medium | 228 | 11,716 | LocalScore ↗ | 512+128 | 113 ms | llamafile 0.9.3 |
| Q4_K - Medium | 226 | 11,437 | LocalScore ↗ | 4,352 | 118 ms | llamafile 0.9.3 |
| Q4_K - Medium | 209 | 11,348 | LocalScore ↗ | 4,352 | 122 ms | llamafile 0.9.3 |
| Q4_K - Medium | 205 | 11,930 | LocalScore ↗ | 512+128 | 113 ms | llamafile 0.9.2 |
| Q4_K - Medium | 198 | 11,534 | LocalScore ↗ | 4,352 | 118 ms | llamafile 0.9.3 |
| Q4_K - Medium | 185 | 11,628 | LocalScore ↗ | 512+128 | 116 ms | llamafile 0.9.2 |
| Q4_K - Medium | 162 | 10,791 | LocalScore ↗ | 512+128 | 126 ms | llamafile 0.9.2 |
| Q4_K - Medium | 161 | 10,794 | LocalScore ↗ | 512+128 | 126 ms | llamafile 0.9.2 |
| Q4_K - Medium | 114 | 10,350 | LocalScore ↗ | 4,352 | 131 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.
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.