Gemini 3.5 Flash at Google I/O 2026: AI that takes the action for you

Tired of AI tools that only answer questions? Google’s bet is that the fix is giving the AI access to your apps.

At its Google I/O conference in May, the company unveiled Gemini 3.5 Flash, a new model that pairs speed with the ability to act across Gmail, Google Photos and other services. This is no longer a chatbot. It is an assistant that takes action on your behalf.

The Flash line has so far been fast but less capable than larger Gemini models. Google now says 3.5 Flash is strong enough to make its own decisions on complex tasks, beating its bigger siblings on several benchmarks.

What “Personal Intelligence” actually means

Personal Intelligence is Google’s answer to Apple Intelligence and Microsoft Copilot. The premise is simple: the AI you use cannot just hold bare knowledge. It must understand you.

In practice: ask Gemini “when do I need to leave for the meeting.” The model checks your Google Calendar. It checks traffic in Maps. It checks Gmail to see whether the meeting is confirmed. Then it gives you a concrete departure time.

All of that with no follow-up prompt from you.

AI Mode in Search with connected apps

Google Search now has an AI Mode that can connect to Gmail, Photos and other apps. You can ask “when was the last time we were at Brioni” and the model will check your geo-tagged photos, find the date and pull up the images.

Privacy, Google says, is handled by keeping data inside your account. Analysis happens in that account’s context. No one else sees what you read or write.

This is the company’s claim, not an independently verified position. Worth keeping in mind.

Speed as the lead advantage against rivals

Gemini 3.5 Flash runs nearly twice as fast as the slowest competing models, such as Claude Opus 4.7 or GPT-5.5 Thinking. On most prompts the difference is invisible because the reply is already quick. On complex multi-step tasks, Flash finishes while the other models are still thinking.

Pricing is also lower. Flash is ten times cheaper than Claude Opus 4.7 per output token. For companies, that means budget for more AI integrations.

What this means for the Google ecosystem

Google is using Gemini to stitch its services together:

• Gmail
• Calendar
• Photos
• Drive
• Maps
• Search
• YouTube

The logic is clean. The more Google services you use, the more useful the model becomes. The more useful it becomes, the more Google services you keep using.

If you already live inside the Google stack, this is good news. If you are on Apple or Microsoft, the open question is how much you want to move your daily life into Google.

Google AI Studio and developer access

Developers can use Gemini 3.5 Flash through Google AI Studio and Vertex AI. The price is three dollars per million output tokens, well below Sonnet or Opus pricing. For teams building their own AI apps, this is an increasingly strong option.

Google is also opening the Gemini API to more regions, which means Serbian developers can hit the models directly without a VPN.

What this means for users in Serbia

A Gemini Pro subscription in Serbia costs around 19.99 dollars per month (about 2,350 dinars) through Google Play. Serbian payment cards work normally. AI Mode in Search is already on for every Google user, regardless of subscription.

Integration with Gmail, Photos and Calendar works automatically. Serbian is supported, although responses on more complex queries are still sometimes better in English.

Is wiring AI into every Google service a good idea?

On practicality, yes. An AI that knows something about you can be much more useful than a generic one.

On privacy, the open question is how much Google keeps and analyzes in the background. Google says it does not train models on your private data. That requires a level of trust some users will not give.

Would you let Gemini read your email and photos for better answers?