mlx-community/mlx-benchmark-leaderboard

🍎 MLX Benchmark V2 Leaderboard

Evaluating LLM proficiency on Apple's [MLX](https://ml-explore.github.io/mlx/) machine learning framework.

520 questions · 11 categories · 6 question types · 4 difficulty levels

Features

• —🏅 Interactive leaderboard with search & column filtering
• —📊 Visual charts — overall comparison, difficulty breakdown, category radar
• —📝 Full documentation of the benchmark methodology
• —📨 Submit your own results to be included

Links

• —Dataset: Goekdeniz-Guelmez/MLX-Benchmark-V2
• —CLI: pip install mlx-benchmark → mlx-bench --model <model>
• —GitHub: Goekdeniz-Guelmez/MLX-Benchmark
• —Paper: Gabliteration (arxiv:2512.18901)

Created by Gökdeniz Gülmez