Skip to main content

​
Overview

A model channel is a stable identifier for a specific AI surface — such as ChatGPT UI, Perplexity API, or Google AI Overview — that stays constant as the underlying model is upgraded or replaced over time. Models change frequently (new versions, new names, API-only models becoming UI models), so filtering by model_id can cause historical data to fragment across several IDs. Filtering by model_channel_id gives you a continuous view of the same surface, regardless of which specific model was powering it on any given day.
Channels are identified by a stable vendor-index string such as openai-0 or perplexity-1. The human-readable description (for example, ChatGPT UI) is returned alongside each channel and may be updated over time.

​
When to use a channel vs. a model

  • Use model_channel_id when you want a stable, long-term view of a surface across model upgrades.
  • Use model_id when you care about the specific model version that produced a response.
A single model may belong to multiple channels when the same underlying model powers more than one surface. To retrieve the list of available channels, use the List Model Channels endpoint.