Contribute model capacity

Release status

Outcome

Connect a device, serve compatible local models, opt in to Network Sharing, and contribute validated work when the production path is released.

Prerequisites

1. A hosted account
2. A released CLI channel
3. A compatible loopback model runtime
4. Explicit consent to the Shared Inference boundary

Current path

1. The released website can sign in an account and approve a one-time CLI device request.
2. CLI installation, local discovery, and provider startup instructions will be published only with the validated release channel.
3. The web account will remain authoritative for enabling Network Sharing per device; the CLI cannot enable it independently.
4. Validation, eligibility, contribution availability, and Credit state will appear only when their authoritative APIs are released.

Authoritative success state

A complete production success state requires released device presence, sharing policy, provider availability, validation, and settlement data. That complete state is not yet available.

Common failures and recovery

1. Keep Network Sharing off for workloads unsuitable for contributor-operated infrastructure.
2. A missing unreleased state is unavailable, not failed or zero.

Security and privacy boundary

The contributor processes plaintext application content. OpenToken shares bounded inference capacity, never model weights, local API keys, or machine access.

Next valid action