Deep scholar (https://deep-scholar.vercel.app, based on Lotus, open-source: https://github.com/lotus-data/lotus) is a semantic search engine coupled with a language model and focused on the search for sicntific articles (Arxiv and PubMed in particular).
Lotus was made by a team of researchers from Stanford and Berkeley Universities (https://lotus-data.github.io/#team).

Once the instruction has been sent, a synthesis is made supported by the summary of the main articles.
Two answers are given: summary and general report.
Answers can sometimes be in English, specify “Generate in French” in the instruction.

The language model used in Deep Research is not specified, but Lotus supports more than 100 models.

Export pdf and share by url.

Free, open-source and unlimited.

Note: The “Sources” tab returns an error during testing (reported).

Test “Generative AI in education, uses and limits”: https://deep-scholar.vercel.app/dashboard/research/828e3e3d-8766-49d9-83c6-3d7c7e229a89

Test “
Search for scientific articles published in 2024 and 2025 that deal with generative AI in education. Search exclusively for scientific articles from sites dedicated to research (google scholar, arxiv, hal, etc.) “: https://deep-scholar.vercel.app/dashboard/research/148cab2e-6909-48f0-bc44-f70efe725ebe