RTX 3080 Ti × Qwen2.5 14B Instruct
Qwen2.5 14B Instruct on RTX 3080 Ti
60.2best gen tok/s
2,090best prompt tok/s
4,352max context tested
7results
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 | 60.2 | 2,014 | LocalScore ↗ | 4,352 | 618 ms | llamafile 0.9.2 |
| Q4_K - Medium | 59.9 | 2,090 | LocalScore ↗ | 512+128 | 617 ms | llamafile 0.9.2 |
| Q4_K - Medium | 58.9 | 1,881 | LocalScore ↗ | 512+128 | 672 ms | llamafile 0.9.2 |
| Q4_K - Medium | 50.0 | 1,959 | LocalScore ↗ | 512+128 | 656 ms | llamafile 0.9.3 |
| Q4_K - Medium | 44.9 | 1,762 | LocalScore ↗ | 4,352 | 708 ms | llamafile 0.9.3 |
| Q4_K - Medium | 36.8 | 1,425 | LocalScore ↗ | 512+128 | 925 ms | llamafile 0.9.2 |
| Q4_K - Medium | 12.2 | 1,420 | LocalScore ↗ | 512+128 | 925 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.
What else can RTX 3080 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.