DeepSeek Starts Billing Search and Fetch Calls at $0.75 per 1,000 on 1 November, Including Those Claude Code and Codex Make on Its API
- What happened
- DeepSeek priced its Search and Fetch APIs at $0.75 per 1,000 successful requests each, and server-side calls from Harness, Claude Code and Codex start billing on 1 November.
- Why it matters
- Coding agents on the DeepSeek API get a new metered line next to tokens, and DeepSeek publishes no per-session request count to budget from.
- What to do
- Before 16:00 UTC on 31 October, measure how many search and fetch requests your agent sessions make on the DeepSeek API.
Our take: the price per request is small; the unknown is volume. If you run DeepSeek Harness, Claude Code or OpenAI Codex against the DeepSeek API, the DeepSeek Search & Fetch API is already in your agent loop, and from 16:00 UTC on 31 October it becomes a metered line on your bill. A coding agent that searches and fetches freely can make many requests per task, and DeepSeek publishes no per-session figure. Count your own requests in the hours they are still free, then decide whether DeepSeek's built-in search stays on. Our directory verdict on the Search & Fetch API stays conditional.
What happened
DeepSeek's change log entry for 17 October opens its in-house search API. The Search & Fetch guide documents two endpoints, both using the same bearer auth as the DeepSeek API:
POST https://api.deepseek.com/v1/tools/searchPOST https://api.deepseek.com/v1/tools/fetch
The guide sets the price: "Search and Fetch are billed per successful request, not by tokens or by the number of returned results." Each costs $0.75 per 1,000 successful requests. A request bills only when it returns HTTP 200 with results; "Failed requests, empty search results, and timeouts are not billed."
Search and Fetch are rate limited independently: 100 queries per second for Search and 25 for Fetch, counted at the account level.
Why it matters
The billing reaches agents you may already run. The guide says: "when you use the DeepSeek API through DeepSeek Harness, Claude Code, or Codex, server-side tool calls already hit the Search and Fetch APIs. These calls are currently free of charge; per-request billing starts at 00:00 Beijing time on November 1, 2026."
00:00 Beijing time on 1 November is 16:00 UTC on 31 October. Nothing changes in how you call the model. A second line simply appears on the bill, next to tokens, and its size depends on how often your agent decides to search or fetch.
What changes for you
- Measure now. Log search and fetch requests per agent session on the DeepSeek API before 16:00 UTC on 31 October, while they cost nothing.
- Budget per request. $0.75 per 1,000 successful Search requests plus $0.75 per 1,000 successful Fetch requests. Failed, empty and timed-out calls are free.
- Watch the Fetch ceiling. Fetch caps at 25 queries per second per account, a quarter of Search's 100, shared across every key on the account.
What we can't tell you
DeepSeek gives no typical request count per session, so we won't estimate a per-task cost. We have not run these endpoints ourselves.
What to do
- 1 If you run DeepSeek Harness, Claude Code or Codex against the DeepSeek API, log search and fetch requests per session now, while they are free.
- 2 Budget $0.75 per 1,000 successful Search requests and $0.75 per 1,000 successful Fetch requests from 00:00 Beijing time on 1 November (16:00 UTC on 31 October).
- 3 Plan around the limits: 100 queries per second for Search and 25 for Fetch, counted per account.
Affected tools & models
Never need to catch up again
The weekly delta: only verdict changes and act-now items. No digest filler.