DeepSeek Launches Web Search and Fetch APIs, Temporarily Lists V4.1 Flash-1017 Snapshot

icon币界网
Share
AI summary iconSummary
DeepSeek has launched new token listings for its Search and Fetch APIs, providing real-time web content for Agent and RAG applications. The APIs cost 5 RMB per 1,000 successful requests and include rate limits. A 'deepseek-v4.1-flash-1017' API briefly appeared before being removed, hinting at a potential model update. Interest rate news remains stable, with no impact on API pricing. The v0.2.1-rc1 version of DeepSeek Harness also introduced new plugin and development features.
CoinDesk reports:

IT Home, October 17 news: This morning, DeepSeek made several changes at the API level.

First, DeepSeek officially announced that the Search and Fetch APIs are now live, both utilizing DeepSeek’s proprietary search engine to provide real-time web content for agents and RAG applications.

The Search API queries web pages and returns structured results, while the Fetch API retrieves a single webpage and returns the title and body as Markdown text. Both are raw JSON APIs with a base URL of api.deepseek.com and use the same Bearer authentication as the DeepSeek API.

According to DeepSeek's documentation, both Search and Fetch are billed based on the number of successful requests, at a rate of 5 yuan per thousand requests. Charges apply only when Search returns a non-empty result list, Fetch returns the page content with an HTTP status code of 200. Failed requests, empty results, and timeouts are not charged. The number of results returned does not affect the pricing.

Each account is limited to 5 queries per second for Search and 5 queries per second for Fetch, counted separately. Fetch also has additional per-domain limits; exceeding these returns HTTP 429. Search requests have a query length limit of 128 characters, with the number of results selectable between 1 and 50, defaulting to 10. Results include a title, URL, summary, a longer model-oriented summary, an optional full text, publication date, re-ranking score, and authority rating.

Notably, searches triggered by Claude Code’s Web Search tool or Codex’s web_search feature will not incur Search fees during the first two weeks after the Search API launch (before November 1, 2026).

At the model level, the bot account AI Tracker Bot on Platform X captured an API entry labeled "deepseek-v4.1-flash-1017" in DeepSeek's official API list, but it was quickly removed.

The account noted that the suffix "1017" is consistent with DeepSeek's previous date-based snapshot naming conventions, such as V4-Flash-0731 and V4-Pro-0813, suggesting it may refer to a new V4.1 Flash snapshot.

The account also updated that the deepseek-flash publicly available on the DeepSeek API remains V4.1 Flash, not the 1017 snapshot; its outputs align with those of third-party-hosted models with the same open weights to a degree comparable to V4 Pro’s consistency across two hosting platforms. As of publication, DeepSeek’s official team has not commented on the existence or changes related to the “1017” snapshot.

As background, DeepSeek officially launched V4.1 Flash on September 10, the smallest model in its new model architecture series, natively supporting multimodal visual understanding. To invoke this model, simply use the name deepseek-flash. The previous versions, V4 Flash and V4 Flash Vision Exp, have been retired and are temporarily routed to V4.1 Flash.

Finally, DeepSeek also released the latest v0.2.1-rc1 version of DeepSeek Harness today, the first release candidate in the 0.2.1 series, which consolidates major user and developer changes since v0.2.0-rc.2. IT Home summarizes the new features as follows:

  • Allow users to discover and install plugins from a custom plugin index source.
  • Added a "Let Agent Create Plugin" entry on the plugin management page, allowing you to save drafts and enter creation mode; execution begins only after submitting the request.
  • Supports enabling, disabling, uninstalling plugins, and synchronizing dependencies after updates; when hot reload is enabled, replaced plugins take effect immediately after update or reinstallation; if hot reload is not supported, a restart is prompted.
  • Optional plugins support minimalist mode; the official plugin list supports on-demand installation of the Claude Code and Codex bundles matched to the current release, and can be used with headless, SDK, and ACP profiles.
  • Added a session state recording interface for experimental plugins to write and read their own session data.
  • Added an experimental compatibility layer for Claude Code Mods. At this stage, the primary goal is to verify that the Claude Code Mods API functionality is largely a subset of the DeepSeek Harness plugin, rather than to provide users with full compatibility.
  • Added an experimental developer tools bundle featuring session raw logs, bidirectional linking between chat groups and messages, and an embedded Node.js debugger; the embedded DevTools currently use an English interface.
  • Added an experimental thought process machine translation plugin, available for enablement on the plugin page, with options to select Bing, Google, or paid DeepSeek Flash.
  • Added an experimental Git Worktrees plugin, available for enablement on the plugins page, which creates and switches to an independent Git checkout by the Agent.
  • The custom model API supports configuring inference levels for each model via the UI, eliminating the need to manually edit configuration files.
  • Pi-ai routing supports updating system prompts within a session based on model capabilities, as well as dynamically adding and removing tools.
  • Compact mode supports image reading and automatically compresses long conversations. Web and desktop users can manually compress context using the /compact command.
  • Global instructions support reading AGENTS.md from the shared agents directory, which can be specified via DSH_AGENTS_HOME.
  • Added the working_directory tool and provided interfaces for reading and switching the session working directory in the TypeScript/Python SDK.
  • When adding a new standalone executable helper runtime, support remote file operations, processes, terminals, sandboxes, and Node PTC.
  • The web version supports logging in to your DeepSeek account when accessing via local HTTP.
  • Supports recognizing keys and tokens in sent messages and defaults to masking them, with optional unmasking as needed.
  • New session support pre-fills unsent prompts; retains files, directories, and session references when restoring drafts or switching workspaces.
  • Add a setting for when to collapse the new workflow step: choose to collapse after the response ends, or when the next regular message arrives.
  • The Markdown file preview displays the YAML header metadata as a clear list of fields.

Related reading:

  • DeepSeek service outage lasted nearly 5 hours; the official compensation covers all API users for consumption within the past 7 days.
  • DeepSeek website/API service has been restored after an outage of nearly 5 hours.
  • DeepSeek website/API services experienced partial outages today, with a message indicating "Server temporarily unavailable."
Source:Show original
Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information. Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.