Model catalog deduplication and canonical visibility hotfix¶

An urgent hotfix synchronizes model catalog representations across the API, Telegram bot, and documentation: eliminating duplicate model entries across endpoints, consolidating pricing attributes, and enforcing canonical key permission checks.

Fixed issues¶

Model catalog row deduplication¶

Fixed an issue where models accessible across multiple endpoint families (such as Chat Completions and Completions) generated duplicate rows in catalog responses.
Consolidated all applicable pricing attributes into a single entry per model, preserving input, output, cache read, cache write, and per-request rates.
Prevented companion endpoints without active pricing rules from appearing in pricing tables.

Canonical model visibility and key permissions¶

Unified model visibility filtering across GET /v1/models, GET /v1beta/models, and Telegram bot screens.
Inactive models and models not granted to the authenticated API key are strictly excluded from public listings and discovery endpoints.
Requests targeting models unauthorized for the key are safely rejected prior to generation.

Model sorting stability¶

Fixed non-deterministic ordering of models sharing the same display priority: model sorting inside each family is now stable and deterministic.

Improvements¶

Cross-surface synchronization¶

The Telegram bot model screens, API model discovery responses, and live documentation tables now share a single canonical source of truth.

No changes to API keys, base URLs, or client configurations are needed.
Check active model availability for your key using GET /v1/models or the Models screen in the Telegram bot.