Model Catalog API

Overview

Model Catalog entries describe the individual LLM models available to the AI Gateway. Each entry carries provider linkage, context/output limits, per-million-token (and per-image / per-audio-minute / per-character) pricing, capability tags, modality and deprecation metadata. The catalog powers model selection, cost accounting and UI listing.

Pricing is not a single flat input/output rate: an entry can also carry a cache-write price queue (pricePerMillionCacheWrite, cacheWritePricesByTtl), long-context price tiers (contextTierPrices) and provider service-tier multipliers (serviceTierPrices) — the gateway applies these only when a request's usage actually reports the matching fact (a cache-write TTL, a prompt above a token threshold, or an echoed service tier). See Create Model Catalog Entry for the full field and sub-object schemas.

capabilities is a normalized tag set, not free text — legacy spellings and modality noise (e.g. chat, json_mode) are rewritten on every save. See Capabilities Normalization.

Endpoints follow the standard AI Gateway APIops conventions:

• Name-based references. The catalog entry is addressed by its name (LiteLLM-style {providerCode}/{modelId}, e.g. openai/gpt-5). Runtime IDs are never required in the path.
• Upsert — POST /{modelName}/ creates the entry, or updates it in place if the name already exists.
• No secret fields — the catalog holds pricing/capability metadata only. It is not a ConnectionConfig subtype, so responses carry no _class discriminator and there is nothing to mask. All fields are returned in plaintext.
• Built-in models are seed-managed — entries with builtIn=true ship with Apinizer. They cannot be deleted, and create/update only applies the price fields (all other fields are preserved from the seed). Custom models (builtIn=false) are fully editable within the owning project.

Endpoints

• List Model Catalog Entries - List all model catalog entries in a project
• Get Model Catalog Entry - Get a single model catalog entry by name
• Create Model Catalog Entry - Create or upsert a model catalog entry
• Update Model Catalog Entry - Update an existing model catalog entry
• Delete Model Catalog Entry - Delete a custom model catalog entry

Base path: /apiops/projects/{projectName}/llm-models

Method	Path	Description
GET	/apiops/projects/{projectName}/llm-models/	List model catalog entries
GET	/apiops/projects/{projectName}/llm-models/{modelName}/	Get a model catalog entry
POST	/apiops/projects/{projectName}/llm-models/{modelName}/	Create / upsert a model catalog entry
PUT	/apiops/projects/{projectName}/llm-models/{modelName}/	Update a model catalog entry
DELETE	/apiops/projects/{projectName}/llm-models/{modelName}/	Delete a model catalog entry

Authentication

All endpoints require authentication using a Personal API Access Token (Authorization: Bearer {token}). JHipster JWTs do not work on APIops.

Permissions

• Reads (GET) require AI_DEVELOPMENT with the VIEW action in the project.
• Writes (POST / PUT / DELETE) require AI_DEVELOPMENT with the MANAGE action in the project.

Related Documentation

• AI Gateway API - Overview of every AI Gateway APIops resource
• Authentication Guide - How to obtain and use API tokens
• Error Handling - Error response formats