Pricing
No subscription. No seat fees. You are charged per token for the AI calls you run, and for nothing else.
The rule is the promise, not the table: you pay per token for the engine tier you point at a review. Rates come from a versioned price book, and the rate charged is stored on every single call, so a past charge never changes because a price did.
The price table
USD per million tokens, by engine tier. These are the same numbers the billing system charges from.
| Tier | Best for | Per 1M in | Per 1M out | Per 100 papers |
|---|---|---|---|---|
| Otter-NanoBig libraries, first-pass triage | Big libraries, first-pass triage | $0.67 | $3.33 | $0.38 |
| Otter-MiniEveryday title and abstract screening | Everyday title and abstract screening | $2.00 | $10.00 | $1.13 |
| Otter-ProFull text, extraction and writing | Full text, extraction and writing | $8.00 | $40.00 | $4.52 |
| Otter-MaxThe hardest judgement calls | The hardest judgement calls | $20.00 | $100.00 | $11.31 |
* Screening only: title and abstract, 4 questions per paper with the second reader on — the same arithmetic the app uses before a run. Full-text extraction sends the whole document and costs more per paper. Your figure moves with your protocol, and the app shows an estimate before every run and the real cost after it.
Signed in, Settings → Billing renders this table live from the billing system, where it cannot drift.
Our model providers can change their prices. When a tier’s price changes we publish a new price book version; work already done keeps the price it was charged at.
What will my review cost?
An estimate, computed with exactly the same arithmetic the app uses before a run.
Estimated cost
$13.71–$15.73
Most likely around $14.58.
This is an estimate, not a quote.
What this assumes
- 8,000 model calls — citations × questions, doubled if the second opinion is on.
- Tokens per call: 282–313 in (typically 301) and 115–134 out (typically 122), for a first-pass screening call. A second-opinion call sends more and writes far less, and is priced separately.
- Title-and-abstract screening. Full-text work sends the whole document, so it costs roughly ten times as much per call.
- Those token figures are measured: the real screening prompt, sent and counted on the sample review. Your abstracts and questions are not those, so treat the result as a close estimate rather than a quote. Once your review has run twenty or so calls, the app estimates from your own history instead.
$50.00 on your first review
$50.00 of AI usage and PDF parsing on us: enough to screen roughly 6,800 abstracts against four questions on Otter-Mini and see the work on your own review.
- Save a card to claim it, once per account and once per card. Saving a card charges nothing. When the credit runs out, load money on the review’s billing page to finish.
- $50.00 in your first review’s wallet, spent the way every charge is: per token, per call, per page.
- Running a larger review? Top it up as you go; there is no plan to choose.
- Your exports are never blocked — not by an empty wallet, not by a payment.
How billing works
- Each review has its own balance. You load money onto it by card ($10 at a time or more) and AI spends from it. Nothing is ever owed, and there is no bill at the end of the month.
- A card belongs to you, not to a workspace. Anyone on a review can attach their card or take it off, so changing who pays is one click. Colleagues see that a card is attached and by whom, never which card.
- Auto-reload is off unless you switch it on. When it is on, the attached card is charged an amount you choose whenever the balance drops below a line you choose.
- You can set a monthly budget that hard-stops AI spending, and a review owner can cap an individual member's monthly spend.
- When a review runs out of money, AI pauses and a running job waits. Load money and it carries on by itself. Your data and your exports are never blocked.
- Every top-up produces a receipt. A statement of money in and usage, and a table of what each citation cost, download from the review’s billing page.
- There are no refunds. Unused money moves to another of your reviews instead.
- A promo code adds a fixed credit to a review, once per person.
- When a model returns something unusable we ask again, and those repair attempts are billed, because the provider bills us. At most 5 attempts per screening answer and 3 per extraction. Connection failures and provider throttling are not billed to you.
What is never charged
Not discounted — not metered at all. There is no meter behind any of these.
- Storage, however many PDFs
- Seats, however many collaborators
- Imports, of any size
- Exports, of any kind
- PDF parsing and text extraction
- Literature searches
You are charged for AI model calls. That is the whole list.
Taxes
Prices are in US dollars and exclude tax. Where we are required to collect sales tax, VAT or GST, it is calculated from the country on your billing profile, added to the charge and shown on the invoice.
Institutional invoicing, purchase orders or larger top-ups: write to support@ottersr.com and we will arrange it.
Three limits, and what each one means
These get confused, so here they are side by side. Only one of them is a limit at all.
| Limit | Measured against | Reaching it means |
|---|---|---|
| Charge threshold | Unbilled balance | Your card is charged. This is normal operation, not a limit. |
| Spend ceiling (2× the threshold) | Unbilled plus reserved | AI waits until the pending charge settles. |
| Monthly budget (yours) | Billed spend since the 1st | AI is refused for the rest of the month. |
| Per-member cap (set by the review owner) | That member's billed spend this month | That member's AI is refused. Everyone else is unaffected. |
Every one of these stops AI only. Screening by hand, validating answers and exporting your data are never blocked — not at the ceiling, not over budget, and not while a payment is overdue.