Tier-gated model catalogue

When you run an AI writing tool, PDF intelligence feature, or logo generation flow, the platform selects from models your plan allows. If a tool requires a higher minimum tier, upgrade paths are explicit rather than mysterious failures.

Teams often mix light and heavy work. Grammar and small rewrites may be fine on lower tiers; long strategic narratives or premium image exploration may justify Elite access. Shared usage means those heavier runs still draw from the same pool — they simply use unlocked catalogue entries.

Because editing is free, humans can always finish the job even when they choose not to spend tokens on another generation. Models assist; they do not become a gate on exporting a deliverable.

Admins should teach “right model for the job.” Unlocking a powerful model does not mean every comment reply needs it. The audit trail will reveal whether the team treats Elite as a default or as a scalpel.

The live catalogue on /ai/models is the reference for what is available. It may include well-known text and image model families delivered through providers. Exact listings change as the marketplace evolves — treat the page as source of truth.

Prompts for AI tools may reach third-party model providers. That is part of how modern AI products work. Read the security page for posture; CasperWasp does not claim SOC 2, ISO 27001, or HIPAA attestation as of the company facts reference.

Transparency also means not inventing fake precision. The catalogue does not pretend to estimate tokens before a run. Afterward, the audit trail records tokens consumed for successful calls.

Design and document tools can expose different model needs. Image generation and long-context analysis are not the same workload as a short tone shift. Tier gating acknowledges that diversity without splitting you into multiple vendors.

Start from workflows, not from brand names alone. If your month is mostly proposal rewriting and simple stamps, a mid tier may suffice. If you constantly need top-end image models and deep document reasoning, price the higher tier honestly.

Use packs when you are allowance-constrained but model-unlocked. Use upgrades when the catalogue blocks the model quality you need. Those are different problems; the platform gives both levers to owners and admins.

Editors should not silently work around tier limits by pasting into consumer AI chat tools — that breaks organisation libraries, audit trails, and shared usage. If the catalogue is insufficient, escalate to an admin for a tier decision.

Revisit tier choice after a real month of audit data. Patterns beat speculation. That is why CasperWasp skips token calculators that invent comfort without your workload.