Skip to content
WebmasterID

Exercise · intermediate

Map a hosted provider relationship

Pick a hosted model in the catalogue, trace creator vs billing provider, and read the hosted pricing reference's source citation.

Estimated time: 10 minutes. Exercises produce evidence artifacts, never model recommendations.

Goal

Demonstrate that you can separate model creator from billing provider for a hosted model, and read both pricing references against their own primary sources.

Prerequisites

  • Read /learn/hosted-vs-first-party so the creator-vs-host distinction is clear.
  • Read /learn/pricing-references so the references-not-quotes framing is in place.

Step-by-step exercise

  1. Open the use case that filters for hosted models

    Visit /use-cases/hosted-inference and read which verified fields the workflow weights.

    Open /use-cases/hosted-inference →

    Expected outcome: You can name the fields the hosted-inference use case asks the catalogue to surface.

  2. Open the hosted-inference selection workspace

    Visit /select?useCase=hosted-inference&hostedAvailability=true. Pick any model in the shortlist.

    Open /select?useCase=hosted-inference&hostedAvailability=true →

    Expected outcome: You have a model slug whose record carries at least one hosted availability row.

  3. Read the model page for creator vs host

    On the model page, identify the providerSlug (creator) and the hosted availability entries (host × hosted model ID × hosted pricing).

    Open /models →

    Expected outcome: You can name the creator and at least one billing platform for the model.

  4. Open the citation for one hosted pricing row

    Click the citation linked to one hosted pricing reference. Note its retrievedAt date and the URL it points at.

    Open /sources →

    Expected outcome: You can quote the host's pricing page URL and the retrieval date.

  5. Note the data gap on the same row

    Look for any field on the hosted record that renders the unverified-data label. If pricing is verified but rate limit is not, that gap is your test plan input.

    Open /learn/hosted-vs-first-party →

    Expected outcome: Your notes name at least one field about the host that you still need to confirm externally.

Completion checklist

Completion checklist

  • You named the model creator and the billing platform separately.
  • You opened the hosted pricing citation directly.
  • You wrote down at least one host-side data gap.
  • You did NOT rank the hosting platforms by price.

Evidence artifact

At the end of this exercise you should have: A /sources citation URL for the hosted pricing reference plus a short note listing the creator, the billing platform, and the data gap.

Paste the artifact into your design doc, ticket, or PR description. The catalogue's role ends with the artifact; the workload-specific testing is yours.

Related workflow routes

Exercise does not recommend a model — external testing still required.

Common mistake

Reusing the model creator's canonical ID in the hosting platform's API call. Each host publishes its own model identifier — the slugs usually differ.

Example artifact

Illustrative example — not a recommendation. Substitute your own values when you run the workflow.

Example artifact

## Hosted-provider mapping (illustrative)
Model creator: <creator-slug>
Host: <platform-slug>
Hosted model ID: <slug-on-host>
Hosted pricing reference: <amount> per <unit> · retrievedAt <date>
Source URL: <host pricing page>

Data gap: <field-name> not stated on host docs — flag for /reverification.

Substitute your real shortlist, slugs, dates, and values when you walk the exercise.

Repeat this exercise when

  • A new hosting platform appears in the catalogue.
  • Hosted pricing rows go stale.
  • The model creator updates the canonical ID or snapshot.

Review before moving on

  • Creator and billing platform are listed separately.
  • The hosted model ID is the one used on the host's API.
  • Hosted pricing retrievedAt is recorded.
  • Any host-side data gap is flagged for /reverification.

Caution: No persistence — checklist resets on every visit. Capture progress in your own notes.

What this exercise does not produce

  • A model recommendation. The exercise routes you through evidence — you decide.
  • A score or grade for any model. The catalogue does not score.
  • A substitute for external prompt, latency, rate-limit, cost, or compliance tests.