Models /
Qwen2.5 0.5B Instruct
5results
3devices
0.63Bparams
300fastest gen tok/s
llama.cpp
| Hardware | Quant | Gen tok/s | Prompt tok/s | Source | Context tested | Build | |
|---|---|---|---|---|---|---|---|
| RTX 3090 | Q6_K | 300 | 19,958 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| GTX 1060 3GB | Q4_K - Medium | 98.7 | 3,114 | LocalScore ↗ | 512+128 | llamafile 0.9.3 | Details → |
| GTX 1660 SUPER | Q4_K - Medium | 63.9 | 1,507 | LocalScore ↗ | 512+128 | llamafile 0.9.3 | Details → |
| GTX 1660 SUPER | Q4_K - Medium | 63.9 | 1,516 | LocalScore ↗ | 512+128 | llamafile 0.9.3 | Details → |
| GTX 1660 SUPER | Q4_K - Medium | 63.8 | 1,520 | LocalScore ↗ | 512+128 | llamafile 0.9.3 | Details → |
Ollama
No measured Ollama results yet. Submit yours →
MLX
No measured MLX results yet. Submit yours →
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.