Public Benchmark Portfolio · §16.1

Quantitative evidence, not vendor claims

We publish 6 benchmarks on a rolling cadence — one per quarter, refreshed when competitive landscape evolution warrants. Each benchmark is methodology-transparent, independently replicable, and executed against AIVAON and named competitors. Results are published with raw data, analysis, and explicit competitive positioning implications.

Latency Benchmark

End-to-end voice response latency across cloud, edge, and on-device runtimes under specified network conditions, measured at p50, p95, and p99 against named competitors.

Maps to attribute A1: Substrate Ownership

Quality Benchmark

Speech recognition and language understanding accuracy across languages, accents, and domain vocabularies, scored against canonical reference sets per language with full per-utterance traces.

Maps to attribute A2: Multi-Jurisdictional Operational Reality

Connectivity Benchmark

Platform behavior under degraded network conditions including intermittent connectivity, high packet loss, and 2G/3G throughput, with completion-rate and recovery-time scoring.

Maps to attribute A4: On-Device Capability

Cost Benchmark

Per-minute all-in cost at specified usage volumes (1k, 10k, 100k, 1M minutes/month) against published competitor pricing, with substrate-attribution breakdown.

Maps to attribute A9: Unit Economic Sustainability

Compliance Benchmark

Regulatory evidence generation time and completeness for specified audit scenarios across GDPR, HIPAA, CCPA, Ghana Data Protection Act, and EU AI Act high-risk classifications.

Maps to attribute A5: Compliance Depth

Sovereignty Benchmark

Deployment topology support and customer control over identity, key material, and data residency. Multi-tenant SaaS, single-tenant cloud, customer-hosted, and air-gapped sovereign topologies are scored independently.

Maps to attribute A7: Identity Sovereignty

Methodology principles

Transparent — every benchmark publishes full methodology, raw data, and the competitive sample so an independent third party can replicate the results end-to-end.
Attribution-honest — competitor results are either measured directly from public endpoints or derived from published specifications with explicit attribution. We never publish a competitor result we cannot point to a source for.
Refreshed — benchmarks are refreshed as the competitive landscape evolves; obsolete results are marked superseded rather than silently replaced.
Charter-anchored — every benchmark measures one or more of the 11 definitional attributes, so the portfolio cumulatively substantiates the category definition rather than producing isolated leaderboards.

Read the Sovereign Voice AI manifesto →