Bunsen is an AI research assistant for molecular discovery, introduced by Schrödinger on July 27, 2026. It aims to interpret research goals and project context presented by researchers in natural language, construct appropriate computational strategies, execute Schrödinger's molecular modeling calculations and multi-stage discovery workflows, and then proceed to result analysis, visualization, and recommendation of the next research steps. While a language model that reads and summarizes paper abstracts can be considered an assistant, Bunsen is closer to a computational chemistry research assistant that translates research questions into actual molecular computation tasks and connects the results back to research decision-making. However, the name of the base model, the internal agent architecture, and the supported individual computation modules cannot be confirmed based solely on the provided discovery materials. In traditional computational chemistry research, researchers must break down questions into individual computation steps, set conditions, and manually compile the results generated from various tools. While general-purpose generative AI is useful for concept explanations or literature summaries, it has limitations in directly planning and executing physical-based molecular simulations and interpreting the resulting trends. Bunsen's differentiation lies in connecting natural language inference with Schrödinger's validated molecular modeling calculations. It goes beyond simply suggesting possible experiments in sentences and integrates the establishment of computational strategies, execution of multi-stage workflows, result analysis, visualization, and recommendation of subsequent steps into a single research flow. This can be understood as an orchestration layer that connects the natural language interface and the domain computation engine in the molecular discovery process. A drug discovery researcher can describe targets, project constraints, a set of candidate molecules, and priority criteria in natural language and use Bunsen to construct multi-stage molecular modeling computation strategies. For example, if a researcher requests that the top 20 candidates be narrowed down from 100 candidates for subsequent review, Bunsen can be used to combine available Schrödinger calculations, compare the results for each candidate, and summarize the observed trends. In the hit-to-lead stage, a virtual workflow can be designed to analyze computational results and structural trends for 50 analogs to prioritize the next cycle of synthesis. The numbers mentioned are example task sizes that researchers can specify, and the actual input limits and processing performance require confirmation from the official documentation. Even when computational results differ from expectations, Bunsen can be used to organize hypotheses and computation sequences for further validation using its result analysis and subsequent research step recommendation functions. Researchers can focus on exploring trends related to research goals and selecting the next computation or experiment candidates instead of reviewing multiple computation results individually. However, quantitative accuracy, computation speed, reproducibility assessment, data security policies, deployment methods, and usage conditions in regulatory environments are not included in the provided materials. Therefore, original computation results and validation procedures should be used in conjunction with critical drug development decision-making, and the conclusions suggested by Bunsen should not be considered as independent experimental evidence.

Bunsen is an AI research assistant for molecular discovery, introduced by Schrödinger on July 27, 2026. It aims to interpret research goals and project context presented by researchers in natural language, construct appropriate computational strategies, execute Schrödinger's molecular modeling calculations and multi-step discovery workflows, and then proceed to result analysis, visualization, and recommendation of the next research steps. If a language model is an assistant that reads paper abstracts and summarizes the key points, then Bunsen is a computational chemistry research assistant that translates research questions into actual molecular computation tasks and connects the results back to research decision-making.

Prioritizing Early-Stage Molecular Candidates: A researcher specifies a target context and 100 initial candidates, along with a scale of 20 for subsequent review, in natural language, and connects Bunsen with Schrödinger molecular modeling calculations to construct a multi-stage evaluation. The resulting trends and top candidates are reviewed to prioritize synthesis and experimental resources.