OpenScholar: An open-source AI tool for literature reviews that outperforms large language models

An open-source AI tool, OpenScholar, has been developed that outperforms major LLMs in scientific literature reviews and accurately cites sources. OpenScholar combines a language model with a database of 45 million open-access articles, directly linking information to its sources to prevent fabricated citations. It is available for free and can be deployed on personal computers. While commercial tools use similar methods, few are open source. OpenScholar is more cost-effective than commercial alternatives like GPT-5 with deep research, though it has limitations in retrieving the most relevant papers.