RTX 4090 × Qwen2.5 14B Instruct
Qwen2.5 14B Instruct on RTX 4090
72.5best gen tok/s
4,361best prompt tok/s
4,352max context tested
13results
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 | 72.5 | 4,361 | LocalScore ↗ | 512+128 | 304 ms | llamafile 0.9.2 |
| Q4_K - Medium | 70.8 | 3,828 | LocalScore ↗ | 512+128 | 335 ms | llamafile 0.9.2 |
| Q4_K - Medium | 69.6 | 4,110 | LocalScore ↗ | 4,352 | 317 ms | llamafile 0.9.2 |
| Q4_K - Medium | 68.8 | 4,208 | LocalScore ↗ | 512+128 | 315 ms | llamafile 0.9.2 |
| Q4_K - Medium | 65.6 | 3,809 | LocalScore ↗ | 512+128 | 343 ms | llamafile 0.9.2 |
| Q4_K - Medium | 60.3 | 4,106 | LocalScore ↗ | 512+128 | 322 ms | llamafile 0.9.2 |
| Q4_K - Medium | 59.7 | 4,122 | LocalScore ↗ | 4,352 | 321 ms | llamafile 0.9.3 |
| Q4_K - Medium | 51.8 | 4,028 | LocalScore ↗ | 512+128 | 334 ms | llamafile 0.9.3 |
| Q4_K - Medium | 48.0 | 3,965 | LocalScore ↗ | 512+128 | 334 ms | llamafile 0.9.3 |
| Q4_K - Medium | 31.9 | 3,161 | LocalScore ↗ | 512+128 | 429 ms | llamafile 0.9.2 |
| Q4_K - Medium | 31.7 | 2,663 | LocalScore ↗ | 512+128 | 473 ms | llamafile 0.9.2 |
| Q4_K - Medium | 16.9 | 2,471 | LocalScore ↗ | 512+128 | 521 ms | llamafile 0.9.2 |
| Q4_K - Medium | 13.1 | 1,973 | LocalScore ↗ | 512+128 | 658 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.