single mention · inference & serving
Agentic search
Agentic search lets a model iteratively invoke search and fetch tools during inference instead of retrieving once.1 Unlike standard RAG, which retrieves context in a fixed pass, it makes multiple tool calls tailored to each query to improve search results.2
Trade-off It outperforms RAG especially on complex tasks at only marginally higher cost.3
How this definition was written
Summarised by glm-5.3-flash from 1 source; every sentence is checked against the quote it cites.
Unlike standard RAG, agentic search empowers the model to iteratively invoke search and fetch tools per query
— DeepSeek-V4: Towards Highly Efficient Million-Token Context IntelligenceUnlike standard RAG, agentic search empowers the model to iteratively invoke search and fetch tools per query, significantly enhancing overall search performance
— DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligenceagentic search being only marginally more expensive than standard RAG
— DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence
Filed under inference & serving › agentic scaffolding