Skip to main content

Release notes in AI Router

2026

September

  • added new metrics to the detailed consumption export in CSV format:

    • cached_input_tokens — number of input tokens read from the cache;
    • cache_write_tokens — number of tokens written to the cache;
    • reasoning_tokens — number of tokens generated by the model during the reasoning stage
August

  • added incoming request masking. If a request contains personal or other sensitive data, the service replaces it with placeholders or rejects the entire request. Masking is automatically applied to existing API methods. There is no need to connect a separate endpoint or modify the request format;

  • improved AI Router consumption viewing:

    • added the ability to view consumption of a specific API key for a month;
    • expanded per-request consumption export to CSV: you can now select the report period and group data by hours, days, months, models, and keys.

    For details, see the guide View AI Router consumption;

  • added the ability to manage limits in AI Router.

June