What the Gemini 3.5 Upgrade Means for NotebookLM
NotebookLM with Gemini 3.5 is Google’s updated, source-grounded AI research notebook that combines advanced reasoning, web source discovery, and exportable outputs in a single workspace designed for complex, multi-step research tasks. The app, originally launched as an experimental tool for understanding documents and topics, now runs on the newer Gemini 3.5 model line, which Google says improves accuracy, reliability, and clarity of reasoning. Internally, the model is paired with more than a hundred curated software skills, giving NotebookLM a larger toolkit for inspecting sources and preparing structured artifacts. According to Google’s own side-by-side evaluations, the upgraded system “achieved an average win rate of more than 65% across five core evaluation categories,” including 69.9% for large document analysis and 78.2% for advanced web research and source discovery, compared with the prior NotebookLM baseline.
Gemini 3.5 Turns NotebookLM into a Smarter Research Partner
Gemini 3.5 does more than speed up answers: it changes how NotebookLM handles research questions from the first prompt. Instead of requiring users to upload documents before it can help, NotebookLM can now start from loose ideas or questions, then use Google Search to suggest relevant web sources that users can selectively add to a notebook. Those chosen materials still ground the AI’s responses, but the setup overhead drops, especially for new projects. The model’s advanced reasoning is visible in the interface: Google promises “better visibility into the thinking process,” with answer-step views tied to Deep Research. For people comparing AI research tools, this moves NotebookLM closer to all-in-one products like Perplexity Spaces or Copilot-style notebooks, while still keeping the focus on traceable, source-based work rather than general chat.
Antigravity and Cloud Code Execution Inside the Notebook
The Antigravity feature adds cloud code execution to NotebookLM, giving each notebook its own secure cloud computer that can write and run code directly inside the research environment. Instead of switching between a notebook, a coding IDE, and a document editor, users can ask a single research question that triggers source selection, calculations, and result generation in one place. Google’s Antigravity 2.0 layer provides “agentic coding capabilities,” meaning NotebookLM can choose tools, inspect material, and run checks as part of its workflow. More than 100 software skills support tasks such as data transformation, validation, and structured analysis. For AI research tools, this is a notable shift: NotebookLM moves beyond summarising documents and becomes a light analysis platform, suited to scenarios like testing hypotheses on tabular data or generating reproducible code-backed research outputs.
Source Discovery and Exports: From Loose Ideas to Finished Artifacts
NotebookLM’s upgraded source discovery and export options are aimed at compressing the full research lifecycle into fewer tools. Discover sources can now reach into the web, suggest relevant material users might have missed, and surface primary sources in other languages or related work by the same author. Once sources are selected and analysed, NotebookLM can turn the results into finished artifacts. It supports PDF reports with charts and tables, .docx documents, Markdown and plain text, plus CSV, JSON and Excel spreadsheets; PowerPoint-style decks, structured data files, charts and images are also available. A studio panel lets users refine outputs after generation, giving more control over instructions and format. These export features help NotebookLM sit alongside, or replace, separate note-taking, spreadsheet, and slide tools for people who need documented, shareable AI-backed research.
Who Gets Access First and How This Repositions NotebookLM
For now, the Gemini 3.5 and Antigravity upgrade is limited to Google AI Ultra subscribers and eligible Workspace customers, putting the most capable version of NotebookLM in the hands of paying individuals and workplace teams before a wider rollout. That access pattern matches how other AI research tools are moving toward premium tiers for advanced features like long-context analysis and code execution. In this release, NotebookLM’s positioning is clearer: it is no longer only a bounded notebook for user-uploaded PDFs and webpages, nor a general chatbot, nor a full coding environment. Instead, it tries to be a comprehensive research and analysis platform that can find sources, inspect them with code, and export traceable outputs. For AI Ultra and Workspace users, this upgrade shifts NotebookLM from helpful side tool to potential central hub for research workflows.
