RTX 3080 × Qwen2.5 14B Instruct
Qwen2.5 14B Instruct on RTX 3080
52.0best gen tok/s
1,741best prompt tok/s
4,352max context tested
6results
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 | 52.0 | 1,595 | LocalScore ↗ | 512+128 | 802 ms | llamafile 0.9.2 |
| Q4_K - Medium | 46.6 | 1,741 | LocalScore ↗ | 512+128 | 734 ms | llamafile 0.9.3 |
| Q4_K - Medium | 41.0 | 1,307 | LocalScore ↗ | 512+128 | 3,576 ms | llamafile 0.9.2 |
| Q4_K - Medium | 31.1 | 874 | LocalScore ↗ | 4,352 | 7,275 ms | llamafile 0.9.3 |
| Q4_K - Medium | 30.7 | 1,169 | LocalScore ↗ | 4,352 | 6,337 ms | llamafile 0.9.3 |
| Q4_K - Medium | 14.4 | 257 | LocalScore ↗ | 4,352 | 11,216 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.