Format benchmark
Measures how language models respond when given the same data in different formats.
The problem
The format you hand a model changes what it costs and how well it answers. This toolkit measures that difference instead of guessing.
- Built with
- Python
- Status
- Open source
- Year
- 2026
What it does
Several formats
The same data is sent as JSON, TOON and other encodings.
Comparable results
Responses are scored the same way for every format.