🇨🇳China·AI News·19 Nov 2026·via Tech Node

    DeepSeek drops separate fees for model-triggered web searches

    DeepSeek has removed separate charges for web searches, image searches, and webpage fetching when its AI models’ built-in tools perform these actions. Effective November 18, developers now pay for these interactions as part of the model tokens used, including the material retrieved and processed.

    DeepSeek has altered its pricing structure, eliminating distinct charges for web searches, image searches, and webpage fetching when these actions are performed through its AI models’ integrated server-side tools. This change, which took effect on November 18, means developers no longer see separate line items for these specific functionalities.

    Previously, these model-triggered actions incurred individual fees. Now, developers pay for these interactions as part of the overall model tokens consumed. This token-based payment covers both the processing by the model and the material retrieved during the search or fetch operations.

    The built-in tools allow developers to create applications that can access current information and read web content as part of an AI model's conversation. The pricing change applies specifically when the AI model itself invokes these integrated tools to perform external data retrieval.

    This adjustment aims to simplify cost calculations for developers, bundling the cost of data retrieval and processing directly into the model's primary usage metric. However, this pricing simplification does not extend to all DeepSeek services. Applications that directly call the company's standalone Search and Fetch APIs will continue to be subject to their corresponding request fees, indicating a targeted adjustment rather than a universal overhaul of all data access charges.

    Share this article

    Go deeper
    Original reporting by Tech NodeWe don't republish, read the full story →

    Related reading

    6 stories