Read the docs
For builders · Product preview

Build around a clear request.

Choose a model that fits the work, review a quote, and keep spending explicit. OFFPEAK is developing an inference flow for applications that can use open models and spare capacity.

Read the docs

Illustrative workflow. Model and service validation are pending.

From a useful task to a checkable record.

Choose the right work.

Start with repeatable text tasks and a measured quality threshold. The pilot is scoped to one approved model revision.

Put a ceiling on each call.

Review an expiring quote and sign a maximum charge. The proposed API makes that limit part of the request.

Keep a record of the result.

Reconcile measured usage and retain the signed billing receipt. Receipt inclusion does not prove correct model execution.