Google unveiled its next-generation artificial intelligence (AI) model strategy for "Gemini 3.5". Gemini 3.5 Flash was released first, but the highly anticipated "Gemini 3.5 Pro" is set to be unveiled next month.

IT outlets Engadget and Business Insider reported on May 19 local time that Google introduced Gemini 3.5 Flash at its annual I/O 2026 developer event in California. It also made it the default AI model for the Gemini app and AI Search mode.

Gemini 3.5 Flash is positioned as a model that emphasizes a balance of speed, cost and performance. Google explained that it was designed with the goal of outperforming Gemini 3.1 Pro in real agent tasks and coding tasks. It is a different model from Flash-Lite and currently aims for faster speeds and lower costs than the Gemini Pro line.

For tasks that require deeper reasoning and longer-context understanding, Gemini 3.5 Pro, scheduled for release next month, will serve as the higher-tier model. Google said it has narrowed the performance gap between Flash and Pro, but Gemini 3.5 Pro has not yet been unveiled.

Google Chief Executive Sundar Pichai (순다르 피차이) said on the I/O stage, "I know a lot of people want to try the Pro model themselves," and asked for time "until next month." He did not explain the specific reason for the delay.

Based on benchmarks presented by Google, Gemini 3.5 Flash scored 76.2 percent on Terminal-Bench 2.1, 83.6 percent in MCP Atlas extension tool usage and 84.2 percent on CharXiv reasoning. Google said it is 4 times faster than major frontier models on tokens output per second.

Google said Gemini 3.5 Flash is suitable for long-running agent tasks. It said the model can reliably carry out multi-step workflows and coding tasks under supervision, and that partners including banks and fintech firms are using it to automate work on a weeks-long basis.

Gemini 3.5 Flash is available through Google Antigravity, the Gemini API in Google AI Studio, Android Studio, the Gemini Enterprise agent platform and Gemini Enterprise. General users can also use it in the Gemini app and in Search AI mode.

The personal AI agent "Gemini Spark" also runs on Gemini 3.5 Flash. Google began distributing it to testers on the day and said it plans to offer an AI agent experience that can run without keeping a laptop open.

Safeguards have also been strengthened. Google said it improved responses to risks related to cyber and chemical, biological, radiological and nuclear (CBRN) threats and lowered the possibility of generating harmful content. It also said it reduced the problem of the model unnecessarily refusing to respond to normal queries.

The announcement is significant in that Google delayed unveiling Gemini 3.5 Pro but deployed the Flash model, which emphasizes speed and cost efficiency, first in real-world use settings. Google is expanding the Gemini ecosystem by spreading the same model family across search, apps, developer tools and enterprise platforms.

A key question going forward is how much of a performance gap Gemini 3.5 Pro will show against Flash when it is unveiled next month. If Flash settles into the role of default model and Pro takes charge of high-level reasoning and long-context processing, Google's AI product strategy is expected to become more clearly centered on separating models by use case.