RTX 4060 Ti × Qwen2.5 14B Instruct
Qwen2.5 14B Instruct on RTX 4060 Ti
27.1best gen tok/s
1,266best prompt tok/s
640max context tested
8results
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 | 27.1 | 1,205 | LocalScore ↗ | 512+128 | 1,154 ms | llamafile 0.9.2 |
| Q4_K - Medium | 26.9 | 1,266 | LocalScore ↗ | 512+128 | 1,074 ms | llamafile 0.9.1 |
| Q4_K - Medium | 26.8 | 1,260 | LocalScore ↗ | 512+128 | 1,070 ms | llamafile 0.9.2 |
| Q4_K - Medium | 26.2 | 1,260 | LocalScore ↗ | 512+128 | 1,093 ms | llamafile 0.9.2 |
| Q4_K - Medium | 24.5 | 1,241 | LocalScore ↗ | 512+128 | 1,093 ms | llamafile 0.9.2 |
| Q4_K - Medium | 24.3 | 1,234 | LocalScore ↗ | 512+128 | 1,102 ms | llamafile 0.9.3 |
| Q4_K - Medium | 23.4 | 1,205 | LocalScore ↗ | 512+128 | 1,122 ms | llamafile 0.9.3 |
| Q4_K - Medium | 3.5 | 327 | LocalScore ↗ | 512+128 | 7,988 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.
What else can RTX 4060 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.