Returns all active LLM models available to the authenticated user's
organization, including full metadata and pricing tiers.

If the organization has no explicit model preferences configured,
all active models are returned (opt-out semantics). Otherwise only
models the admin has enabled are returned.

Typical integration flow

1. Call this endpoint (e.g. ?category=text) and choose a model from models[].name.
2. Either persist it on an agent (PUT /api/agents/{agentId} with model) or pass the same name on POST /api/chat to override for one request.
3. Send chat messages with agentId (and optional per-request model).

Use name, not displayName, in agent and chat payloads. Deprecated or inactive models cannot be assigned and return 400 from those endpoints.