Hardware /

Apple M5 (10-core GPU)

14measured results
101best gen tok/s*
16 GBreported memory

*Headline excludes tiny/test models (under 0.6B params or TinyStories-style). Open in rig checker →

llama.cpp

ModelQuantGen tok/sPrompt tok/sSourceContext testedBuild
Llama 3.2 1B InstructQ4_K - Medium1011,272LocalScore ↗4,352llamafile 0.9.2Details →
Llama 3.2 1B InstructQ4_K - Medium1011,266LocalScore ↗4,352llamafile 0.9.2Details →
Llama 3.2 1B InstructQ4_K - Medium1011,293LocalScore ↗512+128llamafile 0.9.2Details →
Llama 3.2 1B InstructQ4_K - Medium99.21,265LocalScore ↗4,352llamafile 0.9.2Details →
Llama 3.2 1B InstructQ4_K - Medium99.11,272LocalScore ↗512+128llamafile 0.9.2Details →
Llama 3.2 1B InstructQ4_K - Medium97.81,283LocalScore ↗512+128llamafile 0.9.2Details →
Llama 3.2 1B InstructQ4_K - Medium97.11,217LocalScore ↗512+128llamafile 0.9.2Details →
Llama 3.2 1B InstructQ4_K - Medium96.21,250LocalScore ↗512+128llamafile 0.9.2Details →
Meta Llama 3.1 8B InstructQ4_K - Medium22.4211LocalScore ↗512+128llamafile 0.9.2Details →
Meta Llama 3.1 8B InstructQ4_K - Medium22.3211LocalScore ↗512+128llamafile 0.9.2Details →
Meta Llama 3.1 8B InstructQ4_K - Medium22.2210LocalScore ↗512+128llamafile 0.9.2Details →
Meta Llama 3.1 8B InstructQ4_K - Medium19.6177LocalScore ↗4,352llamafile 0.9.2Details →
Qwen2.5 14B InstructQ4_K - Medium11.9111LocalScore ↗512+128llamafile 0.9.2Details →
Qwen2.5 14B InstructQ4_K - Medium11.1109LocalScore ↗512+128llamafile 0.9.2Details →

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.