Upstream token price + 10%
No mystery bundle. No opaque multiplier hidden behind a package name.
Pricing
Provider token cost, published transparently, with a fixed 10% markup.
Persona billing is usage-based. The router selects a provider and model by workload and rigor, and the billed AI price is simply the upstream token rate plus a 10% markup.
Current catalog published from the ai-router service example configuration on April 9, 2026.
Billing formula
provider rate × 1.10
Persona passes through the routed AI token price and adds a fixed 10% markup.
Workload + rigor decide the lane
The router publishes different provider/model lanes depending on the kind of runtime workload Persona is handling.
USD per 1M tokens
All prices below are shown the same way the router config publishes them, so comparisons stay clean.
Current LLM Lanes
The published router lanes are priced by workload, not by one global model.
The current catalog below comes directly from the router example configuration. This is the right thing to publish publicly because Persona does not treat every call as the same workload.
LLM lane
General runtime work
The low-cost fast lane for everyday runtime workloads where the system does not need the heavier reasoning route.
Upstream input
$0.50
per 1M tokens
Upstream output
$3.00
per 1M tokens
Persona input
$0.55
per 1M tokens
Persona output
$3.30
per 1M tokens
LLM lane
Deep runtime analysis
The higher-rigor lane used when those same runtime workloads need a deeper reasoning pass.
Upstream input
$1.75
per 1M tokens
Upstream output
$14.00
per 1M tokens
Persona input
$1.93
per 1M tokens
Persona output
$15.40
per 1M tokens
LLM lane
Planning lane
Planning stays on a dedicated lane across all rigors so this workload keeps a predictable price.
Upstream input
$1.75
per 1M tokens
Upstream output
$14.00
per 1M tokens
Persona input
$1.93
per 1M tokens
Persona output
$15.40
per 1M tokens
Notes
Keep the pricing contract simple.
The published lane can change over time as the router catalog changes. This page is where the current public catalog should live.
The figures shown here are token-based AI pricing only. Any provider-side extras outside token usage should be quoted separately when enabled.
Markup is fixed at 10% over the upstream AI token rate.