Skip to content

Inference under your policies and jurisdiction

Graphene routes workloads under identity-driven residency profiles, so regulated teams keep control of where requests go and every routing decision is auditable. Sovereign and enterprise lanes are sales-assisted and activated under the agreement that matches your lane — not an instant toggle.

Built for demanding operators

Sovereignty is a routing decision, not a consulting engagement. Under an enterprise agreement, workloads tagged to a residency profile are routed only to in-region capacity, and every routing decision is recorded in an auditable ledger.

  • Regulated financial services

    Data-security and cloud obligations (including APRA CPS 234 considerations) make inference residency a board-level question. Sovereign lanes keep prompts and completions in-region.

  • Government and public sector

    Onshore processing with auditable routing for departments and agencies adopting AI under sovereignty mandates.

  • Healthcare and critical infrastructure

    Sensitive-data workloads routed to in-region capacity with policy enforcement at the API layer, not in your application code.

A phased path with programme targets

Figures in the strip below are binding documentation targets from the Graphene corpus — not live dashboards. For public-site data handling, see Privacy.

  1. Design and deploy under your control

    Stand up workspace governance, compliance profiles, and routing rules so the engine receives constraint bundles that match your agreement lane — before widening traffic.
  2. Operate, observe, and scale

    Run production inference with metering and routing visibility; expand regions and supply partnerships when your operating plan and contracts say you are ready.
Engine classify+route p99
< 5 ms
Programme target for the classify-and-route hot path. Not a live latency readout.
Engine total overhead p99
< 15 ms
Programme target versus raw provider latency. Excludes provider time-to-first-token.
Engine availability (Phase 1)
99.9 % rolling 30-day (Phase 1)
Binding programme target for chat/completions paths — not an uptime dashboard.

Regional roadmap

Live means ops-confirmed control-plane availability and contractual data residency for your workspace—not a self-serve toggle on this marketing page. Until your agreement names a region, treat entries below as roadmap posture only.

  • Australia (Sydney)

    Planned

    Status: Planned. First programme geography in architecture docs; marketing reflects roadmap posture — not a live-region claim.

  • Asia-Pacific

    Planned

    Status: Planned. Additional APAC metros follow partner and regulatory sequencing.

  • European Union

    Planned

    Status: Planned. Residency and DPA patterns are agreement-scoped; timelines are roadmap-only until signed.

  • United States

    Not offered

    Status: Not offered. Not positioned for public sovereign marketing in v1; revisit when programme publishes US posture.

Talk to our team

— or send a detailed message below.

Share a few details — we route sovereign and enterprise requests through sales-assisted agreements (see Privacy). No CRM integration in this milestone.