Google Launches Gemini for Science, a Suite of AI Tools for Researchers
Google DeepMind introduced Gemini for Science, a collection of AI-powered tools designed to help researchers work faster across disciplines. The suite combines multiple technologies-Co-Scientist, AlphaEvolve, Empirical Research Assistance (ERA), and NotebookLM-to automate parts of the scientific method that currently consume weeks or months of manual work.
Three experimental prototypes are now available through Google Labs.
Hypothesis Generation
Co-Scientist synthesizes large volumes of scientific literature and generates new research ideas. The system runs what Google calls an "idea tournament," where it generates, debates, evaluates, and verifies hypotheses with supporting citations.
Computational Discovery
Built using AlphaEvolve and ERA, this tool functions as an automated research engine capable of generating and testing thousands of code variations simultaneously. Scientists can explore novel approaches for solar forecasting and epidemiology much faster than traditional methods allow.
Literature Insights
Powered by NotebookLM, this tool helps researchers analyze scientific papers by organizing them into searchable tables and enabling conversational analysis. It can generate reports, slide decks, infographics, and multimedia summaries from curated research collections.
Enterprise Adoption Already Underway
Organizations including BASF and Klarna are already using AlphaEvolve in private preview to optimize supply chains and machine learning models. Daiichi Sankyo, Bayer Crop Science, and U.S. National Labs participating in the Department of Energy's Genesis Mission are using Co-Scientist to accelerate research.
Research papers on ERA and Co-Scientist were published in Nature.
Science Skills Bundle Integrates 30+ Databases
Google also launched Science Skills, a specialized bundle that integrates more than 30 life science databases and tools, including AlphaFold Database, AlphaGenome API, UniProt, and InterPro. The bundle works with Google Antigravity to complete workflows like structural bioinformatics and genomic analysis in minutes instead of hours.
Early internal testing using Science Skills led researchers to identify potential mechanisms tied to a rare genetic disease associated with AK2 gene mutations.
Validation Across Academia
Google is collaborating with more than 100 institutions to validate the systems, including Stanford University, Imperial College London, and The Francis Crick Institute. The company has also partnered with scientific conferences-International Conference on Machine Learning, Symposium on Theory of Computing, and Neural Information Processing Systems-to develop tools for AI-assisted peer review and scientific validation.
Gemini for Science builds on Google's existing scientific tools, which include AlphaFold, AlphaGenome, Google Scholar, Earth Engine, Colab, MedGemma, and Earth AI.
Access to the experiments will roll out gradually through Google Labs.
Related: Learn more about applying AI to scientific research with AI for Science & Research training resources.
Your membership also unlocks:
