Researchers from Sentient, the University of Washington, Princeton University, and UC Berkeley have introduced Open Deep Search (ODS), a new open-source framework designed to match the capabilities of proprietary AI search solutions. The system combines reasoning agents with web search tools to enhance the performance of large language models.
According to the research team led by Salaheddin Alzubi, ODS consists of two main components: Open Search Tool and Open Reasoning Agent. The search tool processes web results more effectively than existing open-source alternatives by rephrasing queries, extracting context from top results, and implementing custom handling for major websites like Wikipedia and ArXiv.
When paired with advanced language models such as DeepSeek-R1, ODS outperforms proprietary services like Perplexity Sonar Reasoning Pro on benchmarks measuring factual accuracy. On the FRAMES evaluation test, ODS achieved 75.3% accuracy, exceeding OpenAI’s GPT-4o Search Preview by 9.7%.
The researchers emphasized ODS’s adaptability, noting that it can work with any base language model chosen by the user. This framework represents a significant step toward democratizing search AI technology, which has been largely dominated by closed commercial systems.
The system offers two versions of its reasoning agent: one based on the ReAct framework, and another using the CodeAct agent. Both approaches enable language models to interpret queries, assess retrieved context, and utilize appropriate tools to generate accurate answers.
The complete open-source implementation has been released on GitHub, inviting developers to build upon and enhance the framework.
Source: VentureBeat