How we measure these numbers
Solve time is measured server-side: the clock starts when our API receives your /solve request and stops when we return a valid token or clearance cookie. It does not include the network round-trip between your machine and our API, which depends on your region — so the latency you observe end-to-end will be slightly higher than the solve time quoted here.
A solve counts as successful only when the token or cf_clearance we return is accepted by the target site's own verification. Failed solves are never billed, which means the success rate and your invoice are measured against exactly the same bar — we have no incentive to count a borderline result as a "solve".
The figures above are aggregated over recent production traffic across all customers, not a one-off lab run on an easy target. We round to honest, conservative values rather than cherry-picking a best case.
Why your results will vary
No solver can promise a single fixed latency, and you should be sceptical of any that does. Your real-world numbers depend on factors that live outside our infrastructure:
- Target difficulty — a lightly-protected Turnstile widget clears far faster than an endpoint running a hardened managed challenge or a fresh Kasada deployment.
- Proxy quality — if you supply a slow or flagged residential proxy, every round-trip through it adds latency and can lower the success rate, regardless of how fast we solve.
- Your region and network — the round-trip to our API from your server adds to wall-clock time; co-locating closer helps.
- Protection churn — Cloudflare and Kasada update frequently. We track and adapt, but immediately after a major change a service can briefly sit at the low end of its success range before our solver catches up.
Why Turnstile is sub-second and Kasada is not
NSLSolver solves on a warm, server-side worker pool instead of cold-starting a headless browser for every request. Turnstile is non-interactive — it runs a small proof-of-work and a set of browser-API probes — so we can produce a valid cf-turnstile-response token directly, which is why Turnstile averages around a quarter of a second.
Cloudflare Challenge and especially Kasada take longer because there is real, unavoidable computation in the way: a full proof-of-work, and for Kasada an extensive browser-fingerprint generation step, must complete before the target will issue clearance. Those seconds are the cost of the protection itself, not overhead on our side — and they are still far faster and cheaper than maintaining your own browser fleet.