Walkthrough
Walkthrough of the five-step decision workflow: pick a use case, build a source-backed shortlist, compare verified fields side by side, surface data gaps, and export a paste-ready evidence brief. WebmasterID Models supports the decision; the reader runs the workload.
What this catalogue is
A source-backed evidence base for AI model selection.
What it is not
An evidence base, not a verdict generator.
Step 1
Filters only mean something inside a use case. Long-context analysis weights context window + pricing; multimodal weights the verified modality channel list; hosted inference weights availability + hosted pricing; governance weights verification status + source freshness.
Step 2
The selection workspace at /select renders candidate models in deterministic shortlist order — verified field count → active lifecycle → source count → name. There is no score, no rank function, no opinion.
Use-case filter narrows the candidate set; provider, lifecycle, min-context, modality, pricing coverage, hosted availability, verification status, and freshness filters tighten it further. Every filtered URL is noindex, follow; the base /select page stays indexable.
Step 3
The comparison builder at /compare/build renders 2–4 selected models against verified fields. No derived metrics, no deltas, no winner. Unknown values render the canonical unverified-data label rather than inventing numbers.
Pre-seed the builder from a use case (long-context · hosted inference · governance review) or pick models manually. Curated comparison pages under /compare follow a higher editorial bar (two-sided verified before indexing).
Step 4
A field marked unverified means the vendor does not publish it (or automated retrieval is blocked) — not that the value is unknowable. The catalogue refuses to guess. Freshness states (fresh / review due / stale / blocked / unknown) are computed deterministically against the build date.
Step 5
The decision-brief builder at /briefs/build renders verified evidence, explicit data gaps, source trail, freshness notes, hosted availability, and a checklist of external tests for 2–4 selected models. Markdown by default; JSON via ?format=json. The brief is evidence, not a recommendation.
Same query shape as the comparison builder. The brief sets X-Robots-Tag: noindex on the export endpoint — generated briefs are team artifacts, not indexable pages. The export endpoint is /api/briefs/decision.
Then — decide what to test externally
The catalogue stops at verified fields. Real selection requires workload-specific testing in your own environment.
Choose a path
The same five-step workflow above, packaged as a curriculum for one audience. Five paths total — beginner, developer, product manager, governance, automation specialist.
Learn first
Each step of this workflow inspects a particular verified field. If the field itself is unfamiliar, the corresponding lesson explains what it means and what it does not guarantee.
How to choose an AI model
The workflow above, in plain language.
Context windows explained
What it means, what it does not guarantee.
Pricing references
Why catalogue prices are references, not quotes.
Hosted vs first-party
Creator vs billing provider.
Model lifecycle
Active, preview, deprecated, retired.
Testing AI models
The work the catalogue cannot do for you.
Test
The AI Usage Lab adds a fourth step to the curriculum. Six testing playbooks (prompt, structured output, long-context, multimodal, automation, regression) and three paste-ready Markdown templates pair with the workflow above. Lab outputs feed straight into the decision brief builder.
Practise
Each exercise routes through one or more of the steps above and ends with a concrete artifact you can share with the team.
Try this workflow
Three pre-built route plans walk the five steps on real verified data. The example decision brief shows what the export looks like without forcing you to build one.