Pricing

There are three ways you pay for StackYak, and they behave differently. Knowing which one a charge comes from tells you when you pay it and what happens when you run out.

Subscriptions

Shared inference is a monthly subscription. You pick a tier, and you can add boosts to it, which appear as extra lines on the same subscription. Plans and tiers covers what the tiers are; Compare tiers carries the prices.

A change to your subscription is charged when it goes through, not at the end of the period.

Prepaid credits

You will be able to buy credit up front and draw it down as you use it.

We are still working out how this works. This page will cover it once the details are settled.

Dedicated compute

A dedicated node is billed monthly, per GPU, on a commitment term of 12, 24, or 36 months. A longer commitment lowers the monthly rate per GPU.

There is no hourly rate on a node and no usage component: you are paying for the hardware for the term, whether or not you send it traffic. Compute pricing carries the rates.

A private endpoint on a stack with a dedicated node is a flat monthly charge per stack. It does not depend on how many machines you enroll or how much traffic they send.

Working out a bill

Each mechanism adds one line:

• Subscription — the monthly price of your tier, plus any boosts on it.
• Prepaid credits — the number of packages you buy, times the package price, paid up front.
• Dedicated compute — the number of GPUs, times the monthly rate for the term you commit to, every month of that term.
• Private endpoint — one flat monthly charge per stack that has one.

Take the rates from the pricing pages and add the lines. Your invoice is built the same way.