Free sample
A free sample of 15 tasks, so you can run your own model against the verifier before deciding anything.
A request form is not possible on this static site. Email us and the sample is sent by email on request.
What to include
- Which families interest you.
- Whether your agent has code execution, since that changes what the tasks measure.
What you get
Task files, the verifier, and instructions for running it. Answers are checked by recomputation; no answer key is included.
After the sample, a paid pilot is optional.