DeepSeek Search and Fetch billing starts November 1, 2026
As of 2026-10-26, DeepSeek Search and Fetch are billed per successful request from 00:00 Beijing time on 2026-11-01. The official Search & Fetch guide says that when you use the DeepSeek API through DeepSeek Harness, Claude Code, or Codex, server-side tool calls already hit Search and Fetch, and these calls are not billed yet.
Updated 2026-10-26
Prices change: on QCode, Claude, GPT and Chinese models are billed per token — live rates on /models.
- Billed per token — live rates on /models
- Pay by card (Visa / Mastercard / AMEX), Apple Pay, Google Pay or crypto
- Self-service: top up and activate instantly
What to check
English docs price
The English docs quote one price for Search and Fetch alike, in US dollars per 1,000 successful requests.
Chinese docs price
The Chinese docs quote one price for Search and Fetch alike, in yuan per 1,000 successful requests. Each page is quoted as written, with no conversion.
QPS limits
Search allows 100 requests per second and Fetch 25, each limited independently; limits apply per account, regardless of API key, and are separate from model concurrency limits.
Billing condition
A request is billable only when it returns HTTP 200 and produces a result; failures, empty search results, and timeouts are not billed.
What changes on November 1
As of 2026-10-26, the official DeepSeek guide says Search and Fetch are billed per successful request from November 1 (Beijing time), not by tokens or by the number of returned results. Server-side calls made through DeepSeek Harness, Claude Code, or Codex already hit these two APIs. The English docs quote the price in US dollars and the Chinese docs in yuan; the numbers are under What to check.
What the official pages say
Details of a successful request: Search must return a non-empty results array, and Fetch must return page content. count only controls how many results are returned and does not change the per-request price. Requests above the account-level QPS limit, or above Fetch's per-domain limit, receive HTTP 429.
Timeline
The English and Chinese versions of the Search & Fetch guide on api-docs.deepseek.com were checked; both say these calls are not billed yet and that per-request billing starts on November 1, Beijing time.
QCode /models lists deepseek-v4-flash, deepseek-v4.1-flash, and deepseek-v4-pro; model calls are billed per token, with per-model rates on /models.
From 00:00 Beijing time, server-side Search and Fetch calls made through DeepSeek Harness, Claude Code, or Codex are billed per request.
Confirmed vs. unconfirmed
Confirmed (verbatim on official pages)
Verbatim in the official guide: Search and Fetch are billed per successful request, not by tokens or returned results; the English and Chinese docs each quote a price; a successful request has an explicit definition; Search and Fetch are limited independently, per account; server-side calls through Harness, Claude Code, and Codex are not billed yet and are billed per request from November 1.
Unconfirmed and not published
The official guide does not say whether calls made before 2026-11-01 will be charged retroactively. It does not say how accounts that call the Search or Fetch API directly, outside Harness, Claude Code, and Codex, are billed before November 1. It also does not mention enterprise accounts, volume discounts, or a spending cap; none of these appear on either language version that was checked.
Per-request billing vs. per-token billing
Per-request billing
Counted by successful requests, regardless of input or output token counts or the number of results returned.
Per-token model billing
DeepSeek model calls are billed per token; the official guide describes only per-request billing for Search and Fetch and does not say how the two are combined.
How to check and prepare
① Open the Search & Fetch guide on api-docs.deepseek.com and read the Billing section of both language versions; each price is quoted as written, so do not convert one into the other. ② Read the Rate Limits section on the same page: Search and Fetch are limited separately, and requests above the account-level limit receive HTTP 429; the guide says to wait and retry on 429. ③ If you use the DeepSeek API through DeepSeek Harness, Claude Code, or Codex, those calls are not billed before November 1 and are billed per request from that day.
DeepSeek models on QCode
As of 2026-10-26, QCode /models lists deepseek-v4-flash, deepseek-v4.1-flash, and deepseek-v4-pro; calls to these models on QCode are billed per token, with per-model rates on /models. Not ready for a monthly plan? Pay-as-you-go Packs start at $4 for a one-time quota you use within 7 days; if you already have a Personal Plan, add a booster pack when you need more quota.
Frequently asked questions
When does DeepSeek Search start charging?
At 00:00 Beijing time on 2026-11-01. The official guide says server-side Search and Fetch calls made through DeepSeek Harness, Claude Code, or Codex are billed per request from that moment.
How much do Search and Fetch cost?
The English docs say $0.75 per 1,000 successful requests; the Chinese docs say 5 yuan per 1,000 successful requests. Search and Fetch share one price, and each version is quoted as written, with no conversion.
Which requests are billed?
Only requests that return HTTP 200 and produce a result are billed. Search must return a non-empty results array and Fetch must return page content; failures, empty results, and timeouts are not billed.
What are the QPS limits?
Search allows 100 requests per second and Fetch 25, per account and regardless of API key. Fetch also has a per-domain QPS limit.
Is there a charge before November 1?
Calls made through DeepSeek Harness, Claude Code, or Codex are not billed yet, and the official guide says per-request billing starts on November 1. It does not say how direct calls to the Search or Fetch API are billed before then.
How are DeepSeek models billed on QCode?
As of 2026-10-26, /models lists deepseek-v4-flash, deepseek-v4.1-flash, and deepseek-v4-pro; model calls are billed per token, with per-model rates on /models.
Sources
The English and Chinese versions of the Search & Fetch guide on api-docs.deepseek.com (fetched 2026-10-26); qcode.cc/models and /pay-as-you-go (fetched 2026-10-26).
Connect DeepSeek models to QCode
Calls to DeepSeek models on QCode are billed per token; rates are on qcode.cc/models.
Related reading
Claude Code Pay-as-you-go Packs: No Subscription, 7 Days
Buy quota once and use it within 7 days, with no subscription; extra packs add to the same key.
DeepSeek Harness + Cordis (2026): Developer Preview Guide
The official DeepSeek agent harness: the Cordis plugin kernel, how to start it, and the mode used for official benchmarks.
DeepSeek V4.1 Flash 1017: released 10-17, context and QCode model
V4.1-Flash-1017 now serves deepseek-flash: 1M context, 384K output, and the model ID to use on QCode.
Crypto payment guide
Nine coins, 56 networks, zero platform fee for AI coding API credits.
DeepSeek billing details on this page were checked on 2026-10-26 against the English and Chinese versions of the Search & Fetch guide on api-docs.deepseek.com; QCode model and billing details were checked on 2026-10-26 against /models and /pay-as-you-go; model availability and rates follow /models. Official changes are not announced separately.