OPENWEIGHT
OPENWEIGHT / A SHORT GUIDE

Intelligence, with a clear path.

Openweight turns verified ownership into a simple way to access curated open-weight intelligence. This guide explains the product sequence at a high level—from entry and eligibility to usage accounting.

A public orientation, not a live workspace.This page describes how the system is designed to work. It does not connect a wallet, run inference, or expose provider accounts, credentials, or infrastructure.
01

Enter the product

Start from Openweight’s entry point and choose to enter the workspace. The public site remains available without an account; product access begins only after the identity flow is completed.

02

Authenticate and connect an EVM wallet

Privy handles authentication and establishes a verified identity. You then connect one canonical EVM wallet—the wallet the system uses when checking eligibility.

IDENTITYPrivy-verified access, rather than a new Openweight password.
WALLETA connected EVM address is the source for the ownership check.
03

Verify configured OPENWEIGHT holdings

Openweight’s backend reads the configured OPENWEIGHT token balance on Robinhood Chain and compares it with the configured threshold. The result is an explicit eligible, ineligible, or temporarily degraded state.

AUTHORITATIVE CHECKVerification happens server-side against the configured chain and token.
NO STALE ACCESSWallet changes or failed checks pause authorization until the state is known again.
04

Eligibility unlocks recurring compute

Eligible holders receive a recurring internal USD compute allowance according to the product policy. Grants are recorded for the relevant period, do not silently duplicate, and do not imply unlimited or transferable compute.

05

Choose a model, then submit usage

Inside the workspace, choose a reviewed, curated open-weight model or select Auto to let the product choose from the available set. When you submit a prompt, Openweight checks access, account state, model availability, and available compute before forwarding the request through its server boundary.

CURATED OR AUTOModel choice stays within the reviewed registry; there is no provider account setup.
USAGEInference requests are metered as they run, including safely handled partial or stopped responses.
06

Reserve, account, and settle compute

Before inference begins, the system reserves an estimated amount from the available allowance so concurrent requests cannot overspend it. Afterward, actual usage is normalized and settled: the reservation is debited for what was used or released when it was not.

RESERVATIONAn atomic hold protects the allowance while a request is in flight.
SETTLEMENTThe ledger records actual usage, releases the remainder, and keeps the balance understandable.