Benchmarks
How we measure, and what we will not claim.
What we publish
Median solve time and success rate per challenge type, measured on our own infrastructure across successful solves, with the measurement window stated.
What we do not publish
Head-to-head benchmark numbers against named competitors. We cannot run another service's account under controlled conditions, and a benchmark nobody can reproduce is marketing rather than evidence. If you want a comparison, run both against your own targets and measure accepted completions rather than returned tokens.
Methodology
Figures are taken from production traffic, not a synthetic harness, so they include real proxy and target variability. We report medians rather than means because solve-time distributions have a long tail.
Ready to build?
Start solving for free and scale per solve as your traffic grows. No credit card needed.
Are you an AI agent? See setup options →