Models /
Deepseek-R1-0528-Qwen3-8B
5results
5devices
8.19Bparams
29.7fastest gen tok/s
llama.cpp
| Hardware | Quant | Gen tok/s | Prompt tok/s | Source | Context tested | Build | |
|---|---|---|---|---|---|---|---|
| RTX 4050 Laptop GPU | Q4_K - Medium | 29.7 | 1,159 | LocalScore ↗ | 512+128 | llamafile 0.9.3 | Details → |
| Apple M1 Max (24-core GPU) | Q6_K | 24.5 | 288 | LocalScore ↗ | 512+128 | llamafile 0.9.3 | Details → |
| Apple M2 Pro (16-core GPU) | Q4_K - Medium | 23.1 | 223 | LocalScore ↗ | 512+128 | llamafile 0.9.3 | Details → |
| AMD Radeon RX 6600 XT | Q6_K | 18.3 | 432 | LocalScore ↗ | 4,352 | llamafile 0.9.3 | Details → |
| AMD Radeon RX 6950 XT | Q8_0 | 5.2 | 1,122 | 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.