About Model Data Points

A data point is one small experiment with AI models: what was tested, on which models, why, and what happened. Each one is a single Markdown file with structured fields, so the entries can be read, filtered and reused.

What these are not

They are not systematic benchmarks. Most are one run on one machine with a handful of tasks. Every entry lists its caveats; read them before drawing a conclusion. For systematic comparisons see Semantic Machines Benchmarks.

Use and licence

The text and data are published under CC BY 4.0; the build scripts are MIT. The entries, raw data files and build script are in the GitHub repository.

Author

Dominik Lukeš, part of Semantic Machines.