Models and Live Catalogue

Primex model availability and pricing are dynamic. Do not rely on a copied list from a guide or an old client cache. Use the live catalogue before building a model picker or selecting a production model.

Customer catalogue

Open the Catalog in the Browser App or Explore → Models in the Telegram Mini App to see enabled and coming-soon models, localized descriptions, billing units, and your effective prices.

For a chat or code model picker, use the authenticated OpenAI-compatible discovery endpoint:

curl https://gateway.primexos.com/v1/models \
  -H "Authorization: Bearer prx-YOUR_KEY"

This request is free and works at zero balance, so compatible clients can load their model picker before a top-up. It returns IDs, not prices or billing units. See Model Discovery.

Public machine-readable catalogue

GET https://app.primexos.com/api/public/catalog is unauthenticated. New clients should consume its sanitized offerings list: stable public IDs, offering type, service/billing metadata, availability, and Primex customer pricing. This public response does not include a legacy models list, supplier prices, or private routing targets.

Choose by capability

Need	Service type or route
Conversation, reasoning, code, and agents	CHAT, CODE, /v1/chat/completions, /v1/responses, /v1/messages
Image and video creation	IMAGE, VIDEO
Speech recognition and synthesis	STT, TTS
Semantic retrieval	EMBEDDING, RERANK
OCR, structured documents, and image analysis	OCR, IMAGE_ANALYSIS
Translation and text analysis	TRANSLATION, LANGUAGE
Content safety	MODERATION
Provider-run web search	WEB_SEARCH usage records; other built-ins are gated

The stable map is described in Capabilities. Exact names, availability, context windows, and rates come from the live catalogue.

Availability states

• Enabled: requests can be served.
• Coming soon: visible for planning, but requests are rejected cleanly.
• Disabled: omitted from customer selection and unavailable to requests.

Cache discovery briefly and handle model_not_found and model_not_available responses. A model being present in an old response does not guarantee it is live forever.

Browse by service type

The sidebar lists auto-generated pages for each service type present in the live catalogue: one page per type with available and coming-soon models, context window, supported protocols, and capabilities. Outcome-based, non-token offerings (OCR, image analysis, translation, text analysis, content safety, and built-in tools) live under AI Services.