How compute units work
Reference for how Uniblock measures, bills, and reports traffic.
What a compute unit is
A compute unit (CU) is a normalized measure of the work behind a single API call. Every endpoint has a published CU cost, surfaced in the pricing calculator and in the dashboard. CUs roll up into a monthly allocation per plan.
Why endpoint costs vary
Endpoints do different amounts of work. A token metadata lookup hits a single cache. A wallet transaction history scans millions of records across multiple providers. CU costs reflect the compute and data cost of each call, not a flat per-request price.
JSON-RPC: weighted averages and ranges
Unified JSON-RPC calls are fanned out across multiple upstream providers. Each provider charges a different CU rate for the same method. Published values are weighted averages computed from production routing logs across all providers in AutoRoute's distribution:
avg CU = Σ (provider_share × provider_CU)
Alongside the average, each method publishes a range — the lowest and highest observed per-provider CU rates during the window. The range conveys routing variance honestly: a call routed to a cheaper upstream costs less than the average, and a call routed to a premium upstream costs more.
Direct API calls bypass routing and pass through to a named provider. Direct endpoints publish a single flat CU — exactly what the upstream provider charges, with no variance.
Why workload profiles matter
A DApp, wallet, and indexer calling the same request volume land on different aggregate CU totals because they hit different method mixes. Traces and logs cost more than reads and balance checks. The calculator's profile selector applies a canonical mix derived from observed traffic patterns so projections reflect realistic workloads.
How plans scale
Each plan includes a monthly CU allocation. Overage rates step down at higher tiers, so larger workloads pay less per overage CU. The Startup plan does not allow overage — traffic beyond the 40M CU allocation returns a rate-limit response until the billing period resets.
| Plan | Monthly | Annual | CU included | Overage |
|---|---|---|---|---|
| Startup | Free | Free | 40M | — |
| Growth | $49 | $529 | 500M | $0.10 / 1M CU |
| Pro | $199 | $2,149 | 2B | $0.08 / 1M CU |
| Business | $549 | $5,929 | 5.5B | $0.06 / 1M CU |
Where actual usage is visible
Per-request CU consumption is logged in real time. The dashboard surfaces it under Analytics → Compute Units, with a breakdown by endpoint, chain, and API key. The numbers there reflect the true CU cost of each call, not the calculator's projection.
How auto-routing affects CU cost
AutoRoute selects a provider per request based on latency, availability, and cost. Each call is billed at that provider's published rate. The average and range shown in the calculator describe the distribution of rates a caller will see across a month of traffic — not a single fixed per-call price.