Pricing

You buy units. A unit is one tenth of a dollar, always shown with its dollar value.

A failed attempt is free. If a change does not pass your build and tests, the whole amount goes back. Try as often as you like within the rules for your plan. Only a change that works and that you accept is charged.
The fast engine goes first. Kintar chooses the right engine for each task. Simple tasks are tried on the cheapest engine that has passed verification. Only a task that fails there moves to a stronger one. Answers are faster and your balance lasts longer.
You see the price before and the receipt after. The most a change can cost is shown before it starts. Afterwards the receipt lists what was charged, what was released and which engines ran.

Units per change

illustrative

1 unit = $0.10, fixed. The exchange rate never changes quietly.

illustrative

Chat: free, on the fast engine, up to 100 messages a day. We cover up to $1.00 of engine cost per account and month; when that is used up, chat pauses until next month.
Code: a free tier of up to 10 attempts (a failed attempt costs nothing but counts toward the 10), at most 3 of them accepted small changes, 7 days to reject a verified change, one free tier per GitHub account and one per repository. No card.
Free capacity is shared and limited: when the free capacity for the day is used up, new free changes and free chat wait, and the page says so. Paid accounts and anything already running are not affected.
Bring your own model keys: planned, not switched on in the app yet. The idea is that you pay your own provider and Kintar supplies the checking.

illustrative

Units are bought in advance. Payments are not switched on yet, so nothing can be bought today.
Cost cap per task: 70 % of the price.
Chat on a paid plan: a message that costs us $0.005 or less is on us; a dearer one is billed at cost, with no markup. Up to 1000 messages a day, and no more than $5.00 of engine cost per account and month on us; past that, chat pauses until the 1st and says so.
A failed attempt is free. If 10 attempts in a row on one repository fail verification within 7 days, new starts on that repository pause until the oldest leaves the window, because each failed attempt still costs us compute.
If fewer than 50 % of the last 20 attempts at one size pass verification, we pause the fast first attempt at that size for 30 minutes (the stronger engine takes over where it can, otherwise the start waits and says so). You are never charged for an attempt that does not pass verification.
Cloud runs: if an account's runs have cost us $10.00 of compute in the last 24 hours, counting runs still going, new cloud starts wait until older runs pass 24 hours. A refused start costs you nothing.
The most a change can cost is shown before it starts.
Computer control is off unless you turn it on. A session is billed by use, not by verified change: the engine and action cost divided by 0.7, rounded up to whole units, never more than the units reserved at the start. A session that ends in an error is not billed.

All figures are illustrative until measured on real tasks. Try the demo.

Size

Price

Cost cap (70 %)

Small

10 units ($1.00)

$0.70

Medium

25 units ($2.50)

$1.75

Large

50 units ($5.00)

$3.50