Priced on cases evaluated.
One meter: every case your suites score. A case counts once however many evaluators judged it, so declaring a fourth criterion does not quadruple your bill.
The rates are not set. The unit is settled and metered in arrears; what a case costs is not decided, so this page quotes no figure. It is not a negotiating position — there is nothing to quote yet.
What a case is
One input scored by one suite, counted once however many evaluators judged it. This is the same number the workspace usage screen shows under Cases evaluated, because a meter and a price that disagree about the unit are two meters.
Systems called and evaluators run are shown beside the bill as the work behind it. They are the honest measure of effort, and what anyone disputing an invoice would ask to see — so they are reported, and the bill does not move with them.
What is not metered
- A rerun that reused its cache. Running an unchanged suite against an unchanged candidate reuses stored executions and judgments and does almost no work. Charging for it would bill for work nobody performed and penalise the caching the product is built on.
- Evaluators that decline the cache. A deterministic check costs less to recompute than to look up, so it misses the cache on every run for the life of a workspace. Those are counted and reported, never charged.
- Reading anything. A workspace that has stopped paying keeps every read. Only new work stops.
Estimate the meter
Not a bill — there are no rates to make one from. This is the volume side, which is the half you can know today: suite size times how often it runs. Find the row nearest yours.
Your own number comes from the engine rather than this table: given a run or a comparison that could not decide, oloproof plan prices the further cases settling it would take, from what that run actually spent.
| Suite size | Merges per week | Cases per week | Cases per month |
|---|---|---|---|
| 500 | 20 | 10,000 | 40,000 |
| 2,000 | 20 | 40,000 | 160,000 |
| 10,000 | 40 | 400,000 | 1,600,000 |
A suite gated on every merge runs as often as you merge.
- Why not price per run?
- Because a run is not a unit of work. One run may score eight cases or eighty thousand, and a repeated run may do nothing at all.
- Why not per evaluator call?
- That was the unit until it was changed on 22 September 2026. It billed a suite declaring four criteria four times over for judging one case once — a bill nobody could predict from the thing they wrote.
- Does a failed run cost anything?
- It costs what it computed. A case the engine did unavoidable work for counts, whatever the verdict was, and a run refused before its first case counts nothing.
- When will the rates be set?
- They are not, and this page will say a number when there is one. Nothing is metered against a rate in the meantime.