{
  "id": 15370343,
  "title": "DeepSeek search and fetch – $0.75 per 1000 successful requests",
  "url": "https://urgent.news/2026/10/18/deepseek-search-and-fetch-0-75-per-1000-successful-requests",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-18T06:50:36.000Z",
  "source": {
    "name": "Hacker News",
    "slug": "hacker-news",
    "url": "https://api-docs.deepseek.com/guides/search/"
  },
  "original_language": "en",
  "account": "The Search API and Fetch API offer access to live web content for agents and Retrieval-Augmented Generation (RAG) applications. They are raw JSON APIs that share the same bearer authentication as DeepSeek's API. To use them, include your API key and JSON content type in the request headers; irrelevant fields in the request body are disregarded. The query is sent as is, including its search syntax.\n\nSuccessful requests yield an HTTP 200 status code along with a results array. If there are no matches, an empty array is returned. Fetching depends on the target site's availability and access policies. Errors adhere to the OpenAI-compatible error format. The param field consistently equals null.\n\nBoth Search and Fetch are read-only operations. HTTP 408, 429, and 5xx errors should be retried after waiting. For HTTP 402 errors, replenish your account balance before retrying. For other 4xx errors, refrain from retrying without modifying the request.\n\nBilled only for successful requests, these APIs do not charge based on tokens or the number of returned results. A request is charged when it returns HTTP 200 and yields a result. However, failed requests, empty search results, and timeouts are not billed. The 'count' parameter determines the number of results returned but does not impact the per-request price.\n\nNote that when utilizing DeepSeek API through DeepSeek Harness, Claude Code, or Codex, server-side tool calls already interact with the Search and Fetch APIs. These calls are presently free of charge; per-request billing commences at 00:00 Beijing time on November 1, 2026. For each account, Search and Fetch have independent quotas, and using one API does not consume the other's quota. QPS (requests per second) limits are determined at the account level, irrespective of the API key used, and are distinct from model concurrency limits. Requests exceeding the account-level QPS limit receive HTTP 429. Moreover, Fetch is also bound to a per-domain QPS limit, and requests surpassing this limit also receive HTTP 429.",
  "summary": null,
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}