DeepSeek opens Search and Fetch APIs at $0.75 per 1,000 calls, with a temporary agent-tool free window

On October 17, 2026 DeepSeek announced that its Search and Fetch APIs are open to all developers. It lists $0.75 per 1,000 calls for each endpoint. Server-side calls made through DeepSeek Harness, Claude Code and Codex are currently free for DeepSeek API users, with billing beginning October 31, 2026 at 16:00 UTC. The announcement coincides with DeepSeek-V4.1-Flash-1017, a new open-weight model revision with claimed coding-agent improvements; that model is related context, not a prerequisite for every search API use. DeepSeek says the underlying search engine serves hundreds of millions of requests daily, but that capacity claim and independent retrieval quality have not been verified.

Web retrieval is a recurring variable cost for coding agents, research workflows and answer systems. A published per-call rate makes DeepSeek a potential alternative to building a search index or relying on a single provider. At the announced list rate, 10,000 Search calls would cost $7.50 before any other model, Fetch, infrastructure or tool charges; 10,000 Search plus 10,000 Fetch calls would cost $15. Developers using the temporary server-side exemption should budget for its expiry and test the actual API's relevance, freshness, content extraction, latency, geographic access, rate limits and acceptable-use conditions before committing production workloads. A provider announcement is not an independent quality benchmark.

DeepSeek's October 17 announcement prices each endpoint at $0.75 per 1,000 calls. A workflow that searches and then fetches a page may incur both types of call. The public announcement does not establish a bundled allowance or guarantee that every result can be fetched. Model token charges, if applicable, are separate from these per-call rates.

DeepSeek says Search and Fetch calls via server-side tools in DeepSeek Harness, Claude Code and Codex are temporarily free for DeepSeek API users. Billing is scheduled to begin at 16:00 UTC on October 31, 2026. This is not described as unlimited free public API access or a permanent waiver; developers should confirm account-specific eligibility and billing before relying on it.

DeepSeek-V4.1-Flash-1017 is available in the DeepSeek app and API, with downloadable weights linked from the release note. DeepSeek claims shorter answers and faster task completion than the prior V4.1 Flash model, using its own benchmark setup. Those claims should not be treated as independently reproduced performance. BTN already covers the underlying V4.1 Flash model line in dossier #84.

Compare query coverage, freshness, citations, page extraction fidelity, failure and retry behaviour, latency and total cost on representative workloads. Separate Search and Fetch counts in telemetry. Revisit costs after the October 31 exemption ends, and check whether the provider's API and billing documentation specify additional limits.