# Ratel Benchmark

> The Ratel benchmark measures how context engineering changes agent accuracy,
> token use, and cost across model families and tool-pool sizes. Results pages are
> generated from the ratel-ai/ratel-bench repository on every build.

## Pages

- [Introduction](https://benchmark.ratel.sh/docs): The Ratel benchmark, measuring how context engineering changes agent accuracy, token use, and cost across model families and tool-pool sizes.
- [BFCL](https://benchmark.ratel.sh/docs/bfcl): Berkeley Function-Calling Leaderboard results for Ratel: task-completion accuracy, token use, and retrieval quality across model families and tool-pool sizes.

## Full text

- [llms-full.txt](https://benchmark.ratel.sh/llms-full.txt): every page concatenated as Markdown