DeepSeek has opened its Search and Fetch APIs to all developers at $0.75 per 1,000 successful requests for each endpoint, according to its API documentation published on 17 October 2026. In the same announcement on X, the company released DeepSeek-V4.1-Flash-1017, an update to its Flash model, with open weights on Hugging Face under the MIT License.
The documentation says that when developers use the DeepSeek API through DeepSeek Harness, Claude Code or Codex, server-side tool calls already hit the Search and Fetch APIs, and that “These calls are currently free of charge; per-request billing starts at 00:00 Beijing time on November 1, 2026.” That is 16:00 UTC on 31 October, our conversion.
Business Pill · THE MODEL AND THE HARNESS
A Business Engineer explainer on the difference between a model and the harness that runs it. It explains the concept in general and says nothing about DeepSeek.
The key insight: As we read it, DeepSeek is pricing a tool the agent calls, not only the model it runs: its documentation lists Search and Fetch at $0.75 per 1,000 successful requests each, and says the calls Claude Code, Codex and DeepSeek Harness make through its API are free until billing starts on 1 November, Beijing time.
DeepSeek describes the Search API as finding web pages for a query and returning structured results, “powered by DeepSeek’s in-house search engine”. The Fetch API retrieves a single web page and returns its content as Markdown. In its X thread, the company says the search engine is “serving hundreds of millions of requests daily worldwide”.
The pricing page lists both APIs at $0.75 per 1,000 successful requests. The documentation says a request is billable only when it returns HTTP 200 and produces a result; failed requests, empty search results and timeouts are not billed, and the number of results returned does not change the price. That works out to $750 per million successful calls on each endpoint, our arithmetic.
Each account is limited to 5 requests per second on Search and 5 on Fetch, counted separately, the documentation states.
Terminal-Bench 4.0 scores as listed in DeepSeek’s own model card for DeepSeek-V4.1-Flash-1017 on Hugging Face (17 October 2026): 42.3 for the new model, 31.2 for the previous V4.1-Flash, 39.9 for GPT-5.6 Sol and 51.8 for Opus 5. DeepSeek says it ran this benchmark with the Claude Code harness. The figures are DeepSeek’s and were not independently reproduced.
What Was Posted on X
๐ DeepSeek-V4.1-Flash-1017 is here!
Stronger benchmark performance, faster task completion, and shorter outputs than the previous versionโa Pareto improvement in performance and cost. Now live on our website, app, and API.
๐ ๏ธ Optimized for @DeepSeekHarness, Tencent CodeBuddy, WorkBuddy, and more. Upgrade to DeepSeek Harness v0.2.1-rc.1 or later for the best experience with the new model.
DeepSeek’s announcement of DeepSeek-V4.1-Flash-1017, posted on 17 October 2026. It describes stronger benchmark performance, faster task completion and shorter outputs than the previous version, says the model is live on its website, app and API, and names DeepSeek Harness, Tencent CodeBuddy and WorkBuddy as tools it is optimized for.
๐ Search & Fetch APIs are now open to everyone!
Powered by our in-house search engine, serving hundreds of millions of requests daily worldwide.
๐ฐ Each endpoint: $0.75 per 1,000 calls. ๐ Docs: https://api-docs.deepseek.com/guides/search ๐ Search & Fetch calls via server-side tools in DeepSeek Harness, Claude Code, and Codex are currently free for DeepSeek API users. Billing starts Oct 31, 2026, at 16:00 UTC.
The third post in the same thread opens the Search and Fetch APIs to everyone at $0.75 per 1,000 calls per endpoint, and says calls made through DeepSeek Harness, Claude Code and Codex are free for DeepSeek API users until billing starts. The piece’s pricing figures come from this post and DeepSeek’s documentation.
The Model Update
The Hugging Face model card describes DeepSeek-V4.1-Flash-1017 as a minor update to DeepSeek-V4.1-Flash that reduces task execution time and output length. In its X post, DeepSeek calls the release “a Pareto improvement in performance and cost” and says it is live on its website, app and API, optimized for DeepSeek Harness, Tencent CodeBuddy and WorkBuddy.
The model card lists DeepSeek’s own benchmark results. On Terminal-Bench 4.0, which DeepSeek says it ran with the Claude Code harness, it reports 42.3 for the new model against 31.2 for the previous Flash version, 39.9 for GPT-5.6 Sol and 51.8 for Opus 5. That is an 11.1-point gain over the previous version, our arithmetic. The scores are DeepSeek’s and have not been independently reproduced in the sources we read.
The API documentation says the model name deepseek-flash is now served by DeepSeek-V4.1-Flash-1017, and that requests to the retired legacy names are served by the new model and billed at the Flash price. The pricing page lists that price at $0.30 per million input tokens on a cache miss and $1.20 per million output tokens at peak, with off-peak rates at half.
Where the Calls Come From
DeepSeek’s documentation says its API is compatible with the OpenAI and Anthropic API formats, and that developers using tools such as Claude Code, GitHub Copilot or OpenCode can use DeepSeek as the backend model directly. It also says DeepSeek Harness is in developer preview for agent harness developers worldwide.
The free period applies to server-side search and fetch calls made through DeepSeek Harness, Claude Code or Codex, according to the documentation. Direct calls to the two endpoints are listed at the $0.75 rate.
As we read it, the free period is tied to the agent tools rather than to direct calls: the documentation says server-side tool calls made through DeepSeek Harness, Claude Code or Codex already hit the two APIs and are free for now, while direct calls are listed at $0.75 per 1,000.
As we read it, the billing rule follows the outcome of a call: DeepSeek says a request is billed only when it returns a result, and that failed requests, empty results and timeouts are not billed.
As we read it, the model release and the search pricing arrive as one package: the same thread that announces DeepSeek-V4.1-Flash-1017 opens the Search and Fetch APIs, and the model card reports its Terminal-Bench 4.0 score with the Claude Code harness.
DeepSeek on X, 17 October 2026
“Powered by our in-house search engine, serving hundreds of millions of requests daily worldwide.”
Three Implications
DEVELOPERS ON AGENT TOOLS DeepSeek’s documentation says search and fetch calls made through DeepSeek Harness, Claude Code or Codex are currently free, with per-request billing from 00:00 Beijing time on 1 November 2026.
BUILDERS CALLING THE APIS DIRECTLY The pricing page lists Search and Fetch at $0.75 per 1,000 successful requests each, and the Search & Fetch guide sets a limit of 5 requests per second per account on each endpoint.
TEAMS ON THE FLASH MODEL The documentation says deepseek-flash is now served by DeepSeek-V4.1-Flash-1017 and that requests to the retired legacy names are billed at the Flash price.
The framework argues that “The surface, not the model, governs autonomy”, and that “Pricing power leaves the model and lands on whoever controls the surface and its reusable workflows.”
As we read it, DeepSeek is placing its search tool inside the surfaces developers already use, Claude Code, Codex and its own DeepSeek Harness, and pricing the tool call itself. The framework was published in July 2026 and does not discuss DeepSeek’s Search API.
What Is Not Established
DeepSeek’s materials do not say how many developers use the Search API, what share of its search traffic comes through agent tools, or what the search engine costs to run.
The benchmark figures come from DeepSeek’s model card. We did not run the model or the APIs, and we did not contact DeepSeek.
Business Engineer Framework
The Agentic Harness War
A Business Engineer analysis of why the agentic surface, the harness, is where value pools in the AI stack.
DeepSeek has opened its Search and Fetch APIs at $0.75 per 1,000 successful requests each, with calls made through DeepSeek Harness, Claude Code or Codex free until billing starts on 1 November, Beijing time, according to its documentation.
It shipped DeepSeek-V4.1-Flash-1017 alongside, with open weights under the MIT License and a Terminal-Bench 4.0 score of 42.3 in its own model card, up from 31.2.
A note on sourcing. This piece is based on DeepSeek’s own materials, read on 17 October 2026: its API documentation for Search and Fetch, its pricing page, the Hugging Face model card for DeepSeek-V4.1-Flash-1017 and its announcement thread on X. Benchmark scores are DeepSeek’s own. The UTC conversion and the per-million and point figures are our arithmetic. We did not run the model or the APIs, and we did not contact DeepSeek. The references to our earlier reporting link to our own pieces.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur, HackerNoon and Search Engine People. Find him on LinkedIn and Substack. See how we source.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
