Pricing

Pay 5% of what we save you.

No plans, no seats, no per-request fee. If we don't save you money, you don't pay us anything.

No credit card, no upfront payment. Your 90 days start with your first routed request, and there's no bill during them. When your 90 days end, you get one invoice covering them: 5% of proven savings, plus applicable taxes and convenience charges. After that, you're billed monthly. You only pay if you want to keep using FewerTokens, and you can walk away anytime. If an invoice isn't paid, routing pauses. Your calls still go through to your default model, with no fee.

No card needed to start.

You only pay when we save you money.

5% of savings

You keep 95% of every dollar we save you, measured against your own default model.

$0 when nothing is saved

Same model as your default: no fee.

How savings are calculated

For every request, we compare two prices for the same tokens:

Cost on defaulttokens × your default model's price

Actual costtokens × routed model's price

Savingscost on default − actual cost

Our fee5% × savings

Your default model: The model you set in the SDK for each provider (for example openai_default_model="gpt-4o"). If we route to a different provider you've connected, we compare against your default for that provider.
List prices: Each provider's published price per 1M tokens, in USD, from our model catalog. Input and output tokens are priced separately.
Never pricier: We never pick a model that costs more than your default, so savings can't go negative.
After your first 90 days, you're billed monthly.

Worked example

Illustrative · catalog list prices

A support bot runs on gpt-4o and uses 30M input + 6M output tokens a month. Suppose 60% of that traffic is simple enough for gpt-4o-mini. The 60% is an assumption for this example, not a measured figure.

gpt-4o $2.50 in · $10.00 out per 1Mgpt-4o-mini $0.15 in · $0.60 out per 1M

Step	Calculation	Amount
Today: everything on gpt-4o30M × $2.50 + 6M × $10.00	30M × $2.50 + 6M × $10.00	$135.00
The simple 60% on gpt-4o18M × $2.50 + 3.6M × $10.00	18M × $2.50 + 3.6M × $10.00	$81.00
The same 60% on gpt-4o-mini18M × $0.15 + 3.6M × $0.60	18M × $0.15 + 3.6M × $0.60	$4.86
Savings$81.00 − $4.86	$81.00 − $4.86	$76.14
FewerTokens fee5% × $76.14	5% × $76.14	$3.81
New LLM bill (paid to OpenAI)$135.00 − $76.14	$135.00 − $76.14	$58.86
New total per month$58.86 + $3.81	$58.86 + $3.81	$62.67 54% lower

Try it with your numbers

Questions

Do I pay anything during the first 90 days?

No bill for your first 90 days. If you stop before they end, you owe nothing. If you want to keep routing, you get one invoice covering them, at 5% of proven savings, plus applicable taxes and convenience charges.

Do I still pay OpenAI, Anthropic or Google?

Yes. Your app calls them with your own keys, so they bill you directly at their normal prices. FewerTokens only bills the 5% fee, plus applicable taxes and convenience charges.

When does the 90-day clock start?

The 90-day clock starts at the first routed request, not at signup. If you choose not to continue, you owe nothing and we simply pause routing and send your calls to your default model, with no fee.

What if my default is already a cheap model?

Then there's little or nothing cheaper to route to, and your fee will be close to $0. Set your default to the model you'd use if FewerTokens didn't exist.

How do I see what I've saved?

Your dashboard lists every request with the default model, the routed model, both costs and the saving.

Is there a rate limit?

Each API key gets up to 120 routed requests a minute. Above that, we send your calls straight to your default model, with no routing and no fee, so you never get an error from us because of the limit. Those calls show as "Sent to your default model" in your dashboard.