Almost every AI tool sells one of three things: a seat, a bucket of credits, or metered usage. The plan names on the pricing page (Free, Plus, Pro, Business) hide which one you are buying, and the difference decides whether your bill is predictable. This guide explains each model with the numbers from our listings, all checked on the vendors’ pricing pages and dated.
Per-seat plans: predictable, sometimes capped
A seat is a fixed monthly price per person. ChatGPT Plus at $20 a month and GitHub Copilot Pro at $10 a month are seats. You know the bill in advance; what you do not always know is the usage cap hiding inside the seat.
Read the fine print on “unlimited”. Most seat plans cap the expensive features: premium model requests, agent runs, deep research reports. Copilot Pro includes 300 premium requests a month; ChatGPT Plus limits the strongest models per window. When you hit the cap, the tool downgrades you to a cheaper model or waits, it does not charge more. That is the trade: predictable cost, variable capacity.
Team seats cost more for governance, not features. Business tiers ($19 to $40 per seat in our listings) add SSO, admin controls, audit logs and the promise not to train on your data. If you need any of those, the extra is cheap; if you do not, the individual plan does the same job.
Credit-based plans: cheap to start, easy to overspend
Credits turn generation into a currency. ElevenLabs sells characters of speech: the $5 Starter plan buys 30,000 credits, the $22 Creator plan 100,000. Runway sells seconds of video: $15 buys 625 credits, and a five-second clip at high resolution can cost 50 of them.
The headline price is a monthly allowance, not a price. The question is not “how much is the plan” but “how many credits does my work consume”. A podcast episode of 40 minutes is roughly 60,000 characters of narration, so the Starter plan runs out in one episode; the Creator plan covers two. Video is harsher: 625 credits is about a minute of finished footage once you count the retries.
Overage is the expensive part. Every credit vendor lets you buy more mid-month at a worse rate than the next tier. If you exceed your plan two months in a row, the upgrade pays for itself.
Usage-based pricing: pay for what you use, if you can measure it
APIs and some developer tools charge per token, per minute or per request with no monthly floor. It is the cheapest model for light or irregular use and the hardest to forecast, because the bill depends on inputs you do not fully control: document length, retries, verbosity of the model.
Set a spending cap on day one. Every API console has one. Without it, a loop in a script or a long document batch turns into a surprise invoice.
Free plans and trials are not the same thing
A free plan (freemium) is permanent with limits: a few file uploads a day, a credit allowance, a watermark. A trial is the paid plan for 7 to 14 days, sometimes with a card on file that converts to a subscription unless you cancel. On our listings, the “Free plan” field shows the actual limits and the “Trial” field says whether a card is required.
Free plans are for evaluation, not for production. Watermarks, non-commercial licenses and low resolution are the usual restrictions. ElevenLabs’ free tier forbids commercial use; Runway’s free credits are a one-time allowance; Descript watermarks exports.
How to estimate a monthly bill in ten minutes
- Name the recurring job: episodes per month, documents per week, minutes of video, lines of code.
- Run one real unit during the trial and note what it consumed: credits, requests, minutes.
- Multiply by your monthly volume, add 30 percent for retries and drafts.
- Pick the plan whose allowance covers that number, on monthly billing first.
- Switch to annual after two or three steady months, when the discount (15 to 35 percent) is worth the lock-in.
Every listing on Webanas shows the entry price, the free plan limits, the trial terms and the date the pricing page was last checked, so you can do this arithmetic before creating an account.
