Notion Adds Gemini 3.5 Flash to Custom Agents with Model Comparison Tool.

Notion announced support for Google's Gemini 3.5 Flash model in Custom Agents, giving users a faster, lower-cost option for building autonomous workflows. The update includes a new comparison feature that lets users evaluate models based on speed, intelligence, and cost before deploying agents.

The addition was announced on LinkedIn on May 29, with Notion highlighting the ability to hover and compare model performance across key metrics. This expands the model options available for Custom Agents, which Notion introduced to help teams automate repetitive workflows by building AI teammates that operate autonomously within workspaces.

Gemini 3.5 Flash is designed as a faster, more cost-effective alternative to larger models. While Notion hasn't published specific pricing or performance benchmarks, the Flash variant typically offers reduced latency and lower token costs compared to standard or Pro versions of Gemini, making it suitable for high-volume tasks like data syncing, content generation, and routine processing.

The model comparison tool addresses a common friction point: choosing the right model for a specific agent workflow. Users can now evaluate tradeoffs before committing to a model, rather than trial-and-error testing after deployment. The interface shows speed, intelligence (presumably accuracy or reasoning capability), and cost side-by-side.

Notion's Custom Agents launched earlier this year as part of the company's broader Developer Platform. As we previously covered, Notion has been actively training users on agent workflows through its AI Academy, offering templates and guides for building agents that handle tasks like ticket triage, content summarization, and database updates.

The Gemini 3.5 Flash integration fits into Notion's strategy of treating agents as core building blocks rather than add-on features. Custom Agents can now tap into tools built with Notion's Workers framework, query external APIs, and continuously sync data from sources like Zendesk or Linear into Notion databases. Adding a cost-optimized model option makes these workflows more practical for teams running high-frequency automations.

For example, a support team using Custom Agents to triage incoming tickets could route simpler queries to Gemini 3.5 Flash for fast categorization, reserving higher-cost models for complex cases requiring deeper reasoning. Similarly, content teams syncing articles or social posts might prioritize speed and volume over nuanced language processing.

The multi-model approach also mirrors trends across Vercel and other platforms building on Notion's infrastructure, where teams have reported 35% faster shipping using Custom Agents integrated with launch databases and automated workflows.

Notion hasn't indicated whether additional models from OpenAI, Anthropic, or other providers will be added to the comparison interface. The current focus appears to be on Google's Gemini family, though Notion's API documentation suggests the platform is model-agnostic by design.

Users can access Gemini 3.5 Flash in Custom Agents starting today. Pricing details and specific performance metrics have not been published in the announcement but are likely available in Notion's agent configuration interface when selecting a model.