Google NotebookLM Upgrade: Gemini 3.5 Flash, Antigravity Cloud Computer, and 65% Win-Rate Performance Gains

Antigravity Integration: NotebookLM Gets a Cloud Computer

The most architecturally significant addition is the embedded Antigravity cloud computing environment. NotebookLM can now write and execute code directly within the notebook context, in service of your research goals.

Google says this comes with a library of more than 100 software skills, enabling users to build automated workflows without switching between applications. Tasks that previously required exporting data, running scripts externally, and reimporting results can now be handled inside the notebook itself.

This is a meaningful shift in what NotebookLM is. It moves the product from a passive research assistant toward an active computational workspace.

Multi-Format Output: Beyond Text

NotebookLM has historically returned text. That changes with this update. The platform can now generate outputs across a structured range of file formats, all surfaced through the Studio Panel alongside existing outputs like audio overviews and quizzes.

Supported formats at launch include:

• Data visualizations and charts — PNG, SVG
• Documents — PDF, DOCX, Markdown, plain text
• Images — PNG, JPG, GIF (via Nano Banana)
• Structured data — CSV, JSON
• Microsoft Excel — XLSX
• Microsoft PowerPoint — PPTX

Crucially, these outputs are editable after generation. Users can prompt NotebookLM to revise a chart, restructure a document, or adjust a dataset — keeping the entire workflow within a single interface. Google has indicated that additional file types will be added over time.

Smarter Web Source Discovery

One of NotebookLM’s defining characteristics has always been its strict reliance on user-provided sources. That constraint has been both a strength — ensuring outputs are grounded — and a friction point when users need to expand their research base quickly.

This update addresses that friction directly. From within the chat interface, users can now ask Gemini to find relevant sources matching their research needs. The model returns a structured research report and offers the option to import all or a selection of those sources into the notebook. All subsequent interactions will draw on those imported sources alongside any manually added ones.

The source-grounded model remains intact. The process of building that source base has simply become faster.

Availability: AI Ultra First

These features are rolling out today, but access is staged. Gemini 3.5 Flash and the accompanying improvements are available first to AI Ultra subscribers, as well as Google Workspace business customers with AI Ultra Access and AI Expanded Access.

Standard Google account holders will receive the updates in the near future. No specific timeline has been provided.

The Bigger Picture

NotebookLM has always occupied a distinct position in the AI tools landscape — purpose-built for source-grounded research rather than open-ended generation. This update does not abandon that identity. It deepens it, while simultaneously expanding the platform’s computational surface area.

The combination of a faster model, embedded code execution, multi-format output, and smarter source discovery makes NotebookLM a more complete research and document automation environment than it was yesterday.

For teams evaluating AI research tools, this update warrants a direct reassessment. The gap between NotebookLM and more general-purpose AI assistants has narrowed — and in several specific dimensions, it may have widened in NotebookLM’s favor.