Vision

B3IQ makes GPU ownership usable at different scales. Customers own the hardware; B3IQ can host and operate it, and an account's applications use that account's machines through standard HTTP endpoints.

The product contract

• Buy a dedicated GPU machine and choose hosting or delivery.
• Manage your machines, runtimes, installed models, keys and SSH access.
• Discover and run models through owner-scoped OpenAI-compatible endpoints.
• Use encrypted private jobs where that transport boundary fits the workload.
• Choose whether idle capacity earns, with estimates and actual records kept distinct.

A missing or unhealthy owned machine does not trigger third-party provider fallback. B3IQ is not a decentralized public inference marketplace. Historical registry, settlement and reputation experiments are not the current product's access or payment contract.

Control you can inspect

Ownership is an access boundary: your gateway key cannot draw from another account's fleet. Readiness is operational: installed models, runtime health, connection, key scope and a successful request each supply different evidence. Planning guidance and benchmark references do not establish current deployment readiness.

Planned capabilities

Serverless-style interfaces, training workflows and verified confidential execution require separate implementation and acceptance evidence. Their presence in a design, guide or roadmap does not make them generally available. The roadmap explains how to evaluate planned work without confusing it with a current purchase option.