Hardware /
AMD Radeon RX 7900 XTX
15measured results
195best gen tok/s*
24 GBreported memory
*Headline excludes tiny/test models (under 0.6B params or TinyStories-style). Open in rig checker →
llama.cpp
| Model | Quant | Gen tok/s | Prompt tok/s | Source | Context tested | Build | |
|---|---|---|---|---|---|---|---|
| Llama 2 7B | Q4_0 | 195 | 3,333 | llama.cpp (Vulkan) ↗ | 512+128 | llama.cpp Vulkan @2f0c2db | Details → |
| Llama 2 7B | Q4_0 | 191 | 3,532 | llama.cpp (Vulkan) ↗ | 512+128 | llama.cpp Vulkan @2f0c2db | Details → |
| Llama 3.2 1B Instruct | Q4_K - Medium | 152 | 4,587 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Llama 3.2 1B Instruct | Q4_K - Medium | 151 | 4,403 | LocalScore ↗ | 512+128 | llamafile 0.9.3 | Details → |
| Llama 3.2 1B Instruct | Q4_K - Medium | 133 | 4,439 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Meta Llama 3.1 8B Instruct | Q4_K - Medium | 51.4 | 853 | LocalScore ↗ | 512+128 | llamafile 0.9.3 | Details → |
| Meta Llama 3.1 8B Instruct | Q4_K - Medium | 51.2 | 888 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Meta Llama 3.1 8B Instruct | Q4_K - Medium | 45.3 | 885 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 32.4 | 489 | LocalScore ↗ | 4,352 | llamafile 0.9.3 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 31.6 | 485 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 27.9 | 502 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Gemma-3-4B-It-Qat | Q8_0 | 20.5 | 1,788 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Llama-3.2-1B-Instruct | Q4_0 | 16.6 | 3,681 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Llama 3.2 1B Instruct | Q4_K - Medium | 16.4 | 3,644 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Llama 3.2 1B Instruct | Q4_K - Medium | 16.0 | 3,734 | 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.