About Crusader Bench
Who Builds It
@hesamsheikh
Creator and maintainer
Builds and runs Crusader Bench: the harness, the agent, the dataset and this site. Crusader Bench is an independent project, not affiliated with Firefly Studios or with any AI lab.
Contribute
Crusader Bench is open source, and help is welcome: fixes to the harness and its tools, new maps, and the work on the roadmap. Anyone with the game can also play games on their own setup and submit them to the dataset as a pull request, where they are shown as community games.
Help Test More Models
Every game is paid for in API credits. Labs that want their models on the leaderboard, anyone who wants to support the project and anyone with a question are welcome to get in touch.
The Benchmark in Numbers
- Models tested
- 7
- Games published
- 28
- Episodes played
- 34
- Tokens used
- 124.6M