DeepSeek launches Search and Fetch APIs, new Flash model snapshot spotted

DeepSeek has rolled out several API updates. Its new Search and Fetch APIs are now live, both powered by its own search engine. They bring real-time web content to Agent and RAG apps. Search returns structured results for a keyword, while Fetch grabs a single page and returns it with the title and body.

Both APIs charge only for successful requests, at 5 yuan per 1,000 calls. Failed calls, empty results and timeouts are free. Search can return 1 to 50 results per query, with title, URL, summary and authority score included.

On the model side, a tracker account briefly spotted a deepseek-v4.1-flash-1017 entry in DeepSeek's official API list before it was removed. Based on DeepSeek's date-based snapshot naming, it may point to a new V4.1 Flash snapshot. The company has not commented yet.

DeepSeek also released Harness v0.2.1-rc1, adding a custom plugin index source, plugin hot reloading, a thinking-process translation plugin and more.