Benchmarks
Submitting results
Run the public cases through a tool, upload what it returns, and we score it - credited to you, published after review.
Anyone with an account can add a tool to the leaderboard. You do not need to own the tool.
- Open the benchmark, for example PDF, and download the cases. Use Download all to get one zip.
- Run each case through the tool exactly as a user would. Record the plan or tier you used.
- Go to Submit results, choose the tool and how you ran it (web app, desktop app or API), and upload the files the tool returned, unchanged. Files named after their case are matched automatically.
- Select Send for scoring.
We score every file ourselves with the same scorer the CLI uses. Scoring takes a minute or two per case. After scoring, an editor reviews the submission, and it joins the leaderboard once approved. You can follow its progress, and read any review note, on your submissions.
What gets refused
- A file that is not a PDF, or is larger than 20 MB.
- A file whose page count is not the case's page count.
- A tool that is not in the catalog. Submit it first.
You do not have to cover every case. The leaderboard shows how many cases a result covers, so a tool tested on one case is never read as tested on all of them.