What a CAPTCHA price change actually does to your cost per 1,000
Every few months a solving provider changes its rates, and the same question goes round the same channels: is this still the cheapest option? It is the wrong question, and it is wrong in a way that costs money.
The two models
Per-solve billing. You pay for each solve. Cost rises in a straight line with volume: double the scraping, double the bill. The useful property is that it is proportional — small months are cheap. The expensive property is that a rate change applies to every solve you will ever run, and you feel it immediately at the top of your volume.
Thread-based billing. You pay for concurrency — how many CAPTCHAs you can have in flight at the same moment. Solves inside that capacity are unlimited. The useful property is that your bill stops tracking your volume: once the threads are paid for, more solves through the same threads cost nothing extra. The expensive property is that you pay for the peak even in the hours you are idle.
These are not two prices for the same thing. They are two different shapes — and the shape matters more than the rate.
The arithmetic
Take one plan and hold it still — ADVANCE, $90 a month, 50 threads, solves unlimited within those threads. The monthly bill does not move, so cost per 1,000 is simply the bill divided by what you push through it:
| Solves per month | Cost per 1,000 |
|---|---|
| 10,000 | $9.00 |
| 100,000 | $0.90 |
| 500,000 | $0.18 |
| 1,000,000 | $0.09 |
Nothing clever is happening here. It is one number divided by a bigger number. But it is the whole argument: on per-solve billing your cost per 1,000 is a constant, and on thread-based billing it is a falling curve. Comparing a per-solve rate against a plan price without saying at what volume is comparing a constant to a curve and picking a winner at random.
It also explains why a per-solve rate change lands so hard. A rate rise moves the whole line up at every volume at once. There is no volume you can grow into where it stops applying.
The ceiling is real — work from your peak, not from the maximum
The falling curve has a floor, and anyone selling you thread-based pricing should tell you where it is.
One thread handles one in-flight CAPTCHA. It frees when that solve finishes. So throughput per thread per hour is:
# solves per hour, per thread 3600 ÷ seconds_per_solve # reCAPTCHA v2, ~65s median on our own daily tests 3600 ÷ 65 ≈ 55 solves / hour / thread ≈ 40,000 / month (at full utilisation)
reCAPTCHA v2 takes around 65 seconds on our own daily tests — it is an interactive challenge, not an OCR call, and most of that time is the challenge itself.
Nobody runs at full utilisation. Traffic arrives in bursts, jobs finish, retries cluster. So that 40,000 is a ceiling to reason about, not a number to plan with. Size your threads to your peak concurrency — the most solves you ever have in flight at once — and treat the per-thread maximum as headroom you will not reach. Faster types move the arithmetic in your favour; slower interactive ones move it back. Which is why the honest version of this calculation uses your mix, not an average.
Where thread-based billing is the wrong answer
You run low or bursty volume
At 20,000 solves a month, per-solve billing is almost certainly cheaper for you, and you should stay on it. The curve above has not fallen far enough at that volume to beat a per-solve rate, and paying a monthly minimum to solve 20,000 CAPTCHAs is paying for capacity you are not using.
You need concurrency far out of proportion to volume
If your workload needs 100 threads for a fifteen-minute burst and almost nothing for the rest of the day, you are buying a peak you occupy briefly. Per-solve billing handles that shape better, because it charges you for the work and not for the shape of it.
The crossover between the two is a real number and it is specific to you. It depends on your volume, your per-1,000 rate, and the concurrency your workload actually needs.
Work it out for your own numbers
Do not take the table above as an answer — it uses one plan at four volumes, and your mix is not that. Put in the rate you pay today and the volume you actually run, and read the crossover.
Open the cost calculator Start a 3-day trial
The calculator does not ask you to believe anything about anyone's pricing. You enter what you pay; it does the division. Plan prices above are CaptchaAI list prices; the 65-second figure is the median of our own daily reCAPTCHA v2 tests (n=14, August 2026).
