Google releases Gemini 3.5 Flash and autonomous agent Gemini Spark at I/O 2026

Google announced two significant AI releases at its I/O 2026 conference this week: Gemini 3.5 Flash, which now powers the company's AI Mode search feature, and Gemini Spark, a cloud-based autonomous agent designed to handle tasks across multiple services.

Gemini 3.5 Flash powers enhanced search

Gemini 3.5 Flash is powering Google's AI Mode, which allows users to ask follow-up questions and corrections during searches. The model supports multimodal inputs including images, video files, and entire Chrome tabs as direct search inputs. According to Google, the AI-powered search box uses AI to "anticipate your intent and help you formulate questions" rather than simply autocompleting text.

Gemini Spark: autonomous task execution

Gemini Spark represents Google's entry into autonomous AI agents. Running entirely in the cloud, Spark can monitor credit card statements for hidden subscriptions, track updates from school emails, and compile notes into Google Docs. The agent can interact with third-party applications including OpenTable and Instacart to complete tasks.

Google claims Spark will request user confirmation before making purchases or sending emails, addressing concerns about autonomous agent control.

Gemini Omni generates video from any input

Google also announced Gemini Omni, described as a generative AI model that can "create anything from any input." Gemini Omni Flash is rolling out to the Gemini app, Google Flow, and YouTube Shorts. The model accepts images, audio, video, and text as input to generate videos. According to Google, Omni "better understands physical forces such as gravity, kinetic energy and fluid dynamics" compared to previous models like Veo 3.1.

New pricing tiers

Google introduced a new $100-per-month "AI Ultra Plan" positioned between its existing $20 Pro plan and top-tier Ultra plan (reduced from $250). The mid-tier plan offers five times higher usage limits than Pro, priority access to Google's Antigravity coding tool, and 20TB of cloud storage. The top-tier plan provides 20 times higher usage limits and exclusive access to Project Genie, which allows users to build interactive 3D worlds using Google Street View imagery.

Pricing details for API access to Gemini 3.5 Flash were not disclosed.

What this means

Google's release of Gemini Spark signals a direct move into autonomous agent territory, competing with emerging startups and potentially OpenAI's reported agent development. The integration with third-party services like OpenTable and Instacart suggests Google is building an ecosystem approach rather than a standalone agent. The $100/month pricing tier indicates Google is segmenting its AI offerings for power users who need higher usage limits but don't require experimental features, potentially capturing customers who find $20/month insufficient but $250/month excessive.