Models /
Gemma 3 1b It
9results
5devices
1.30Bparams
199fastest gen tok/s
llama.cpp
| Hardware | Quant | Gen tok/s | Prompt tok/s | Source | Context tested | Build | |
|---|---|---|---|---|---|---|---|
| RTX 3090 | Q4_K - Medium | 199 | 17,009 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| GTX 1080 Ti | Q4_K - Medium | 129 | 5,556 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| GTX 1080 Ti | Q4_K - Medium | 128 | 5,367 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| RTX 4050 Laptop GPU | Q4_K - Medium | 98.1 | 8,031 | LocalScore ↗ | 512+128 | llamafile 0.9.3 | Details → |
| RTX 4050 Laptop GPU | Q4_K - Medium | 87.0 | 8,007 | LocalScore ↗ | 512+128 | llamafile 0.9.3 | Details → |
| RTX 3060 Laptop GPU | Q4_K - Medium | 73.8 | 6,477 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| AMD Radeon RX 6650 XT | Q4_K - Medium | 60.7 | 3,528 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| RTX 3060 Laptop GPU | BF16 | 22.1 | 3,121 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| RTX 3060 Laptop GPU | BF16 | 21.0 | 2,679 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | 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.