DeepSeek launches Search and Fetch APIs at prices up to nine times cheaper than rivals

The company introduced the APIs on October 17 alongside its DeepSeek-V4.1-Flash-1017 model. The new model reportedly scored 77.0 on the DeepSWE v1.1 benchmark, ahead of Claude Opus 5, which scored 74.0, according to the figures cited in the announcement.

The pricing strategy could significantly reduce the cost of building AI agents that rely on live web data to research topics, monitor developments and generate responses.

According to DeepSeek’s pricing page, Search and Fetch each cost $0.75 per 1,000 successful requests. Failed requests, empty results and timeouts are not billed.

Charges will begin at midnight Beijing time on November 1, equivalent to 16:00 UTC on October 31. Until then, calls through DeepSeek Harness, Claude Code and Codex will remain free.

The pricing applies globally, without regional differences, while each endpoint is limited to five requests per second per account.

DeepSeek undercuts established competitors

The new rates are substantially lower than those of competing retrieval services.

Brave Search API charges $5 per 1,000 requests, while Exa charges $7 for standard search and $12 for deep search. Tavily's basic search pricing ranges from $5 to $8 per 1,000 requests, depending on the plan. Perplexity Sonar starts at $5 per 1,000 requests, with additional token costs.

At $0.75 per 1,000 requests, DeepSeek's service is around 85% cheaper than providers charging $5 and nearly 94% cheaper than Exa's $12 deep-search tier.

The launch also marks DeepSeek's first move into charging for tools beyond its token-based model pricing.

The company has built its reputation on comparatively low inference costs. Its Flash model has offered input tokens at $0.15 per million during off-peak periods, with cache-hit pricing as low as $0.003 per million tokens.

Backed by the resources of High-Flyer, the Chinese quantitative investment firm, DeepSeek has attracted considerable attention across the open-weight AI ecosystem. Its new APIs extend its low-cost strategy into the retrieval infrastructure that AI agents need to access external information.

Lower costs for AI agent developers

The price difference could be particularly significant for developers operating agents that repeatedly search the web and retrieve information.

An agent completing 10,000 search-and-fetch cycles daily could incur retrieval charges of $7.50 per day at DeepSeek's advertised rate, compared with $50 to $120 using competing services charging $5 to $12 per 1,000 requests.

Actual costs will depend on how many individual Search and Fetch requests each cycle generates, as the two endpoints are billed separately.

The move comes as other parts of the AI infrastructure market become increasingly competitive. Companies such as Modal Labs have attracted substantial investment by providing orchestration services above increasingly commoditised computing infrastructure.

DeepSeek's entry into retrieval could intensify competition in another part of the AI agent technology stack, putting pressure on established providers to justify their higher prices through differences in search quality, reliability, features or service limits.

Regulatory questions and long-term pricing

China's Ministry of Commerce and Ministry of Science and Technology issued Announcement No. 48 of 2026 on October 16, proposing export controls covering 23 AI technologies.

The announcement's reported exemptions for API and software-as-a-service access could allow overseas developers to consume certain Chinese AI services without the licensing requirements that may apply to exporting or redistributing model technologies. The precise scope of the proposed rules and their application to individual services will depend on the final regulatory framework.

DeepSeek's free access period also gives developers an opportunity to integrate the APIs before billing begins on November 1. Early adoption could encourage developers to build their retrieval workflows around the company's services, although switching costs will depend on how easily those systems can be adapted to competing providers.

The long-term sustainability of the pricing remains uncertain. DeepSeek reportedly operates a search engine that handles hundreds of millions of requests daily, potentially allowing it to benefit from economies of scale. However, the company does not publish audited financial statements that would establish its cost per request or the profitability of the new services.

The impact will become clearer once paid usage begins and developers gain more experience with the APIs' performance, reliability and total operating costs.

For now, DeepSeek's pricing introduces a substantially lower-cost option for AI agent retrieval and could force competing providers to reconsider their pricing or offer clearer differentiation.