Google NotebookLM Boosts Research with Gemini 3.5 and Code Execution

Google has rolled out a significant upgrade to its AI-powered research assistant, NotebookLM, integrating its advanced Gemini 3.5 model and the Antigravity coding tool. This update aims to deliver substantial improvements in information accuracy, reliability, and transparency regarding the AI's reasoning process. The core enhancement includes embedding a "secure cloud computer" within each notebook, allowing the system to write and execute code. This capability facilitates deeper research and more complex data analysis directly within the platform, moving beyond mere summarization.

TheThe upgraded NotebookLM now boasts over 100 specialized software skills, dramatically expanding its functionality. According to Google's internal evaluations, the new system achieved an average win rate exceeding 65% across its top five core evaluation dimensions, marking a 15 percentage point improvement over its previous iteration. Specifically, it demonstrated a 69.9% win rate in large document analysis and an impressive 78.2% win rate in advanced web research and source discovery, indicating enhanced performance in handling extensive information and locating relevant sources.

A key benefit for users is the expanded range of output formats. NotebookLM can now generate documents, data visualizations, and structured data in various professional formats, including PDFs, DOCX files, Microsoft Excel (XLSX), PowerPoint (PPTX), CSV, JSON, and multiple image types like PNG and JPG. This allows users to create ready-to-download reports, spreadsheets, and presentations directly from their research, streamlining workflows and reducing the need to switch between multiple applications for final output generation.

This strategic upgrade positions NotebookLM as a more comprehensive and versatile tool for researchers, analysts, and anyone dealing with large volumes of information. By enhancing its analytical power, transparency, and output flexibility, Google aims to make complex information processing more accessible and efficient for an international audience, enabling users to build source libraries from initial ideas and maintain full control over the included and cited materials.