Skip to content

Pricing

Token-based metering

You pay for the input units and output units your function consumes per compute call, multiplied by the function's cost_multiplier.

Playground

Free

Hero playground on the homepage. Demo tenant, capped at 30 calls per IP per hour.

  • Built-in functions only
  • No credit card
  • Shared rate limit
Try it

Dev

Pay per inference

Your own tenant, your own API keys. Pay only for the metered units your functions consume.

  • Built-in + custom functions
  • JWT + scoped API keys
  • Per-tenant rate limits
  • Webhooks (HMAC-SHA256)
Open app

Marketplace provider

Revenue share

List your function and earn per inference. Revenue is recorded automatically per call.

  • Public marketplace listing
  • Auto revenue recording
  • Monthly payout periods
Browse marketplace

Cost multiplier explained

Heavier functions weight more per call. Built-in multipliers:

Functioncost_multiplier
Jale Price Optimizer2×
Cost-Plus (no incumbent price required)1×
Competitive Price with Business Floor1×
Optimal Stopping Rule1×
Jale Elasticity Calculator1×
Jale Advanced Recommendation5×
Elo A/B Test Engine1×
Jale Psychological Pricing1×
Rosetta Translate3×
Rosetta Normalize2×
Rosetta Validate1×
Rosetta Suggest2×
Rosetta Freeze1×
Rosetta Ingest3×
Elo Significance1×
Elo Bandit Allocator2×
Jale Bundle Pricing3×
Jale Tier Optimizer3×

Worked example: 142 input + 38 output units at cost_per_unit = $0.005 × cost_multiplier = 2× → $0.002560. Computed via the same computeMeteringCost helper the inference router uses.

How we price the pricer

Three things worth knowing before you rely on a price from us.

Is the base rate computed?
No. The rate per unit is a number we set by hand. It is not learned from your data or from the market. We could price our own pricing with a pricing function, but that function would also need a price, and so on without end. So we picked one number, and we tell you it is picked.
Does acting on a price change what the next price learns from?
Yes. Learned pricing estimates demand from the quotes your workspace made and whether buyers accepted them. Once you act on a price, the market you measured has moved. If you only ever quote prices close to the current one, the data says little about prices further away. Treat a learned price as an estimate that can go stale, not a fixed truth.
Why is the cheapest route not always the best one?
The cheapest strategy picks the function with the lowest billed rate per call. That rate is what the call costs to run. It says nothing about whether the price that comes back is a good one, and a good price can be worth far more than the saving. Cheapest suits high-volume work where cost is the limit. For decisions with money at stake, pick a route on other grounds.