Skip to main content

Custom tokenizer depending on provider

Feature Request

Token estimates seems to depends heavily to openai api.

As far as I know, each model provider (openrouter, gemini, openai, …) has it own method to retreive the token counts for each request (for example, openrouter use await fetch("https://openrouter.ai/api/v1/generation?id=$GENERATION_ID", { headers })

Status: Closed1 comment

Log in to comment and vote

Comments1

  • Tony Dinh changed status to Closed
    Team•

    Mar 2, 2024

    Unfortunately there isn't a reliable way to count the exact token usage.

    I think you should check the dashboard of the provider to be sure.