Introduction
B3IQ turns local inference hardware into manageable, verifiable AI compute nodes with OpenAI-compatible endpoints and protocol accountability.
B3IQ is a pilot network. Local inference, hosted gateway foundations, real routed traffic, signed receipts, customer usage rows, node payout rows, and payout settlement pilots exist. Production network claims still depend on provisioning and update trust, broader durable supply, production gateway controls, abuse controls, and recurring protocol automation.
In the press
What B3IQ does
Run a local node Your provisioned machine runs the host agent: it inventories hardware, serves the setup wizard, manages local API keys, and exposes local OpenAI-compatible inference.
Call standard endpoints Use HTTP directly. B3IQ keeps SDKs and adapters out of scope so the contract stays endpoint-first and portable.
Understand route readiness See why a node or model is not earning yet: heartbeat, route, runtime, benchmark, queue, and customer policy all have explicit blockers.
Trace protocol evidence Public surfaces expose safe commitments such as receipt hashes, payout roots, readiness status, and reputation inputs without raw prompts or outputs.
Get a machine Buy a dedicated GPU server that B3IQ racks and operates, or enroll hardware you already own. Either way, your machines serve your OpenAI-compatible traffic free.
Product surface
Host agent Linux Go service, supervised by systemd, for identity, inventory, local APIs, model jobs, receipts, heartbeats, and command polling.
Wizard React setup flow for claiming a node, choosing mode, creating local auth, installing models, and checking readiness.
Gateway Hosted OpenAI-compatible model and chat endpoints backed by router policy, own-account routing, and receipts.
Protocol Node registry and registration/onboarding state, payout settlement commitments, public explorer views, and future reputation roots.
Network signals
Explorer surface
