The CAPTCHA Solver Price War Has a Number Now
With Turnstile and DataDome standard on every Tier-1 site, solver economics has become a P&L line item. CapSolver's AI-only model undercuts 2Captcha's hybrid human pool by 50%+ on Turnstile while claiming ~99% success.
For most of the last decade, CAPTCHA solver pricing was a footnote in scraping budgets — a few cents per 1,000 reCAPTCHA v2 puzzles, irrelevant against bandwidth and compute. That changed when Cloudflare Turnstile, AWS WAF Bot Control, and DataDome’s slider went mainstream. Solver costs are now a real line item, and the price gap between vendors has widened to the point where it shows up on the balance sheet.
Here is what the solver market actually charges in Q1 2026.
CapSolver runs Cloudflare Turnstile at roughly $1.20 per 1,000 solves. 2Captcha charges $2.89 per 1,000 for the same — a 140% premium. On reCAPTCHA v2 image grids, AntiCaptcha is still the cheapest at around $0.50 per 1,000 if you can tolerate slower turnaround. On DataDome’s slider challenge, CapSolver runs roughly $2.00 per 1,000 with a claimed success rate near 99%.
The 2x–3x price spread between vendors on the same challenge type is the new reality. It is not because one vendor is incompetent. It is because the solver business has bifurcated.
The bifurcation
CapSolver and the newer entrants run AI-only solving — no human workers in the loop, just trained models. That keeps marginal cost low and lets them compete on price for high-volume, structurally similar challenges (Turnstile, DataDome slider, reCAPTCHA v3).
2Captcha and AntiCaptcha keep human worker pools, with median solve times around 13 seconds on reCAPTCHA / Turnstile. That overhead shows up in the price, but it pays for coverage of the long tail: 2Captcha covers more than 30 distinct CAPTCHA types including AWS WAF, hCaptcha Enterprise, and GeeTest v4. The AI-only vendors do not.
For a publisher running a single-target scraper hitting one CAPTCHA type, the choice is straightforward: take the cheapest vendor that passes that specific challenge. For a publisher running a multi-target operation, the math is messier. You either pick one vendor that covers everything (and pay the premium), or route per-challenge through multiple vendors (and pay the integration overhead).
Apify Store actor publishers fall mostly into the first category. The actors with measurable demand on the Apify lead-extractors segment — the ones documented in the Q1 2026 census — overwhelmingly target LinkedIn, where the relevant challenge is Cloudflare Turnstile sitting in front of authentication. CapSolver wins that segment on price.
The Cloudflare-induced floor
The other thing that changed in 2025 is that Cloudflare added a browser-validation challenge in February 2025 that fingerprints real hardware via a Chrome bug. Many DOM-only solver flows that previously worked on Turnstile silently broke. The vendors that adapted shipped solvers that combine the Turnstile API call with stealth-browser hardware emulation; the vendors that did not lost share.
The practical effect is that the cheap-and-fast solver-as-API model is no longer sufficient for the highest-value targets. To clear Cloudflare reliably in 2026 you need: (a) a Turnstile solver, (b) a stealth browser stack capable of passing the hardware check, (c) a residential or mobile proxy with appropriate IP reputation, and (d) cookie persistence across sessions. The solver is one line item in a four-line budget. Or you can route through a managed unblocker like Bright Data’s Web Unlocker, which bundles all four primitives into a single per-request cost (roughly $3 per 1,000 successful requests).
What this means for actor publishers
For Apify Store publishers running PPE pricing, the implication is concrete: the per-event price floor on a CAPTCHA-protected target is now visibly above zero. A LinkedIn-targeting actor that hits Turnstile on every run owes roughly $0.0012 per record to CapSolver before accounting for proxy, compute, or margin. An actor priced at $0.001 per record cannot survive that math. An actor priced at $0.005 per record can.
The leaders in the lead-extractors segment — harvestapi, dev_fusion, apimaestro — all price above the solver floor. The long-tail “spray” publishers like contacts-api (210 actors averaging 2 users each) often do not, which is part of why their economics do not work.
The open downstream question is what happens to the solver vendors when pay-per-crawl becomes widespread. If the publisher-set crawl fee is lower than the cost of solving the CAPTCHA, the solver gets bypassed entirely. If it is higher, the solver stays in business. CapSolver’s $1.20 per 1,000 sets one bound. Cloudflare’s pay-per-crawl rates — once they are public — will set the other.
For now, the solver market has finally produced public, comparable, per-call pricing across the major challenges. That is more legibility than this segment has ever had. It is also a sign that the fight is no longer about whether scrapers can clear protections — it is about which P&L line absorbs the cost.
Sources