NeuroAI NEUROAINEUROAI.SITE
China AI, reported clearly
1,300+ verified stories on models, robotics, BCI and the business of AI.
Browse stories
ESC

DeepSeek drops separate fees for model-triggered search and web fetch

DeepSeek removed separate fees for model-triggered search and web fetch, while model tokens still bill and direct Search/Fetch calls retain per-request pricing.

2026-11-19 · 164 words · NeuroAI
DeepSeek drops separate fees for model-triggered search and web fetch

DeepSeek has removed separate charges for search and web fetch when the model itself decides to invoke them, while keeping billing for the underlying tokens and retaining per-request pricing for direct calls to the Search and Fetch endpoints.

The change looks small and is not. Under the old scheme, an agentic query that triggered retrieval produced a bill with three components, which made cost forecasting difficult for developers building assistants.

Folding retrieval into token billing makes the price of a task roughly proportional to what the model writes and reads - a simpler unit for product teams, and one that discourages prompt patterns that trigger excessive tool calls.

It also nudges behaviour: if retrieval is effectively free at the margin, developers are more likely to enable it by default, which improves answer freshness at the cost of higher token consumption.

Competitors will have to respond on structure, not just headline price, because the differentiator for agent platforms is increasingly how predictable the bill is.

Related coverage

Cite this article

NeuroAI (脑机网), "DeepSeek drops separate fees for model-triggered search and web fetch," 19 November 2026, https://neuroai.site/a/na-auto-2026-11-19-deepseek-search-fees

Machine-readable: BibTeX · RIS · JSON · Markdown

More in “Foundation Models” → · Back to home · Markdown version

China AI, reported clearly
1,300+ verified stories on models, robotics, BCI and the business of AI.
Browse stories