Skip to content
AI Technology News

Opinion2 min read

DeepSeek adds Search and Fetch APIs at $0.75 per 1,000 successful requests

DeepSeek now offers a Search API and a Fetch API for agents and RAG apps, billed per successful request instead of per token. Server-side tool calls are free until November 1, 2026.

By Ola-obaado Oluwadara · Updated Oct 19, 2026, 7:20 AM UTC

On this page

DeepSeek has added two web-access endpoints to its API: Search and Fetch. Search takes a query and returns structured results from DeepSeek's own search engine. Fetch takes one URL and gives you the page back as Markdown. Both are plain JSON over POST and use the same bearer-token auth as the rest of the DeepSeek API.

Search supports operators like site:, -site:, exact phrases, lang:, after: and before:. You can ask for 1 to 50 results, and set need_content to get the full page text rather than just a snippet and a longer model-ready summary. Each result also carries a relevance score (0 to 3) and an authority tier, so you can filter out weak sources.

Pricing is $0.75 per 1,000 successful requests for each API, not per token and not per result returned. A request only counts if it returns HTTP 200 with something in it, so failed calls, empty searches and timeouts are free. Limits are 100 requests per second for Search and 25 for Fetch, per account, plus a per-domain limit on Fetch. If you reach DeepSeek through DeepSeek Harness, Claude Code or Codex, the server-side tool calls are free for now. Billing starts at 00:00 Beijing time on November 1, 2026.

My take

The pricing model is the part I like. Paying only for successful requests means a dead URL or an empty search doesn't cost you, and that matters for agents, which hit broken links all the time. Flat per-request pricing is also easy to predict. You can work out a monthly bill before you ship.

The Fetch API returning clean Markdown is practical too. Anyone who has built a RAG pipeline knows how much time goes into stripping HTML, nav bars and cookie banners. If this works well, it removes a whole chunk of glue code. But the docs say fetching depends on the target site's access policies, and a 404 covers many failures, so I'd want to see how it handles JavaScript-heavy pages and blocked sites before trusting it.

My suggestion: don't rip out your current search setup yet. The first things I'd check are result quality outside English and Chinese queries, and how fresh the results are. The docs don't say how DeepSeek's index compares with other search APIs, so any claim about quality is unproven. Test it on your own queries.

For builders in Nigeria and elsewhere in Africa, a per-request price in dollars is simple to budget, though you still need a way to pay in dollars and top up the balance (a 402 error means you ran out). Also note the free period ends November 1, 2026. If you're prototyping through a tool that calls these endpoints server-side, plan for the cost before that date, not after.

Sources

This post was written with AI assistance from the sources above.

Written by

Ola-obaado OluwadaraSoftware Enigeer

Fullstack software engineer skilled in building scalable web applications, implementing secure systems, and deploying solutions to the cloud. Experienced in both frontend and backend development, with a strong focus on performance, reliability, and clean code.

GitHubLinkedInX

Get the weekly AI roundup

The tools worth your time, tested by a developer. One email a week.

One email a week, no spam. Unsubscribe any time. See the privacy policy.

Need AI features built for your business?

I build practical AI into products: assistants, search, automation.

Contact me

Comments

What do you think? Drop your thoughts below.

Be kind and stay on topic. Comments with links are checked before they appear.

Loading comments…