Free again after 17 days of charging: Deep Search cancels DS model network search and content crawling fees

📅 2026-11-19

Abstract:

Deep Search has recently adjusted the billing rules for the DS API networking function: starting from 12:00 on November 18, 2026, server-side web search, image search and web crawling triggered by the model will no longer be charged based on the number of calls, and only the costs related to model input and output will be calculated.

Deep Search previously billed server-side tools and independent search and crawling APIs from November 1st. Approximately 17 days after the new billing rules are implemented, the model will be automatically connected to the Internet and the free mechanism will be restored. Developers who directly call the search interface will continue to be billed based on the number of calls.

Old version of search crawling billing rules (implemented on November 1st):

  • Search request fee

  • Web scraping costs

  • The cost of model processing content

Search and crawling are calculated based on 5 yuan per 1,000 successful requests. For example, if a certain task searches once and reads 3 web pages, the tool call fee is 0.02 yuan, and the Token consumed by the model to generate answers is billed separately. After the adjustment on November 18, if search and web page reading are triggered by the server-side tool of the model, this part of the 0.02 yuan tool fee will be cancelled, and only the basic Token processing fee will remain.

Type 3 server networking methods will be free again (implemented on November 18):

  • Web search, image search and reading links in DeepSeek networking tools

  • Anthropic compatible web_search in the API, including Claude Code Web Search

  • Responses API web_search

  • DeepSeek Harness Networking operations when selecting DeepSeek legacy search provider

In order to facilitate developers to count the behavior of the model when processing tasks, the API response will still return the search and crawl times through server_tool_use, but these server-side tool calls will no longer be charged accordingly, and developers can simply ignore this part.

The independent search and crawling interface is still charged on a per-use basis:

It should be noted that when developers directly call the independent search and crawling interface through the API, they will still be charged on a per-time basis. The search interface needs to return HTTP 200 and the list is not empty before it will be charged, and the crawling interface will only be charged after successfully obtaining the web page body. There will be no billing for failed requests, no search results, or crawling timeouts.

Therefore, if developers only need models to find and read information by themselves, they can use server-side networking tools to reduce costs. When you need to control search results, sorting, or RAG processing yourself, use a paid independent search and crawling interface.

Related tags

Related articles

Comments

0/500
Captcha (click to refresh)
No comments yet