DeepSeek removes separate fees for model-triggered web search and fetching

DeepSeek stopped charging separately for server-side web search and fetching triggered during model calls on November 18, according to its official pricing documentation. The change took effect at noon Beijing time, or 04:00 UTC.
Model token usage is still billed. The change covers server-side search invoked through DeepSeek’s web tools and the web_search tool in its Anthropic-compatible and Responses APIs.
This is not a blanket removal of retrieval charges. Standalone Search API and Fetch API calls remain listed at $0.75 per 1,000 successful requests each. Developers therefore need to distinguish a model’s integrated web operations from direct calls to the separate retrieval endpoints.
The pricing page also continues to list file parsing and optical character recognition at $0.225 per 1,000 pages. Those services are separate from the search-and-fetch change.
The update removes an additional billing component from model-driven browsing workflows. It does not make those workflows entirely free: the model’s input and output token charges still apply, and direct retrieval calls retain their own rates.
Sources and citations
Revision history
Published .