Gemini 3.5 Flash: Faster, smarter, and safer model

The new Gemini 3.5 Flash model becomes the default model in the Gemini app and in Google Search's AI mode starting May 19. According to official Google materials, it is significantly faster than other top-tier AI models and brings the ability to generate "richer and more interactive" web user interfaces and graphics.

A key strength of Gemini 3.5 Flash is its performance in agentic tasks — scenarios where an AI model independently performs multiple steps, connects to external tools, or decides on the next course of action. Google claims that in agentic and coding benchmarks it even outperforms its older sibling Gemini 3.1 Pro, which is remarkable given that Flash has traditionally been the "lightweight" and cheaper variant of the Pro model.

Another important update is the improved safety guardrails. According to Google, Gemini 3.5 Flash should be less prone to generating harmful content while also less frequently mislabeling harmless queries as risky. This is crucial for companies that want to deploy AI in production environments without worrying about uncontrolled outputs.

Gemini 3.5 Flash is the first model from the new Gemini 3.5 family. The Pro version will arrive next month and can be expected to bring even higher performance for more demanding tasks.

Neural Expressive: A complete visual transformation of the app

Along with the new model comes a redesign of the entire Gemini app, which Google internally calls Neural Expressive. There are several changes:

• Smoother animations and a more vibrant color palette
• New typography designed to improve answer readability
• Haptic feedback when touching buttons in the mobile app
• Reformatted answers — the most important information is now at the top, supplemented with images, interactive timelines, narrated videos, and dynamic graphics
• Easier switching between text input and Gemini Live voice mode

The redesign is gradually rolling out on the web version of Gemini and in the Android and iOS apps starting May 19. Additionally, Gemini Live will soon gain support for regional dialects, which could be interesting for Czech users — the question remains whether Czech will appear among the supported languages.

Gemini Spark: Google responds to OpenClaw

One of the most interesting announcements is Gemini Spark — an always-on AI agent that works in the background while you do something else. Google is responding here to the OpenClaw phenomenon, the open-source AI agent platform that took the tech world by storm earlier this year.

Gemini Spark is powered by the Gemini 3.5 Flash model and runs on Google Cloud virtual machines 24 hours a day, 7 days a week. What can it actually do?

• Send emails and reply to messages
• Browse credit card statements and find hidden subscription fees
• Create summaries of meeting notes
• Connect to Workspace apps (Docs, Sheets, Slides) and third-party tools like Canva, Instacart, or OpenTable
• Work with local files on macOS through the Gemini app

Gemini Spark opens to selected testers this week. Google AI Ultra subscribers in the US will have beta access starting next week. European availability has not yet been announced, but given that Google typically rolls out new features in the EU with a delay of several months, we can expect arrival during the summer or autumn.

Gemini Omni Flash: On-demand video from anything

Google is also launching an entirely new model family called Gemini Omni. The first of these, Omni Flash, can generate video clips — not just from text like the previous Veo model, but from any combination of inputs: text, photos, video, and audio.

Google promises realistic-looking scenes with more accurate physics and the ability to "think" about what happens next in the video. It also includes a feature called Avatars, which lets you create videos with a digital version of yourself using your own voice. The output can then be edited through conversation directly in the Gemini app.

Gemini Omni Flash is available to Google AI Plus, Pro, and Ultra subscribers. It can be used for free in YouTube Shorts and the YouTube Create app.

Daily Brief: Your personal AI assistant for a daily overview

Another new agent is Gemini Daily Brief, which connects apps linked to Gemini and creates a summary of your day. It pulls upcoming events from Calendar, important messages from Gmail, and organizes everything according to your priorities and goals.

The system learns from your feedback — just give the summary a thumbs up or thumbs down. Daily Brief is available to US Google AI Plus, Pro, and Ultra subscribers starting May 19.

Premium plan prices drop

The Google AI Ultra subscription has also undergone a significant change. It drops from the original $249.99 per month (roughly CZK 5,700) to $100 per month (about CZK 2,300), matching OpenAI's ChatGPT Pro subscription.

A new $200 per month tier has also been added, which includes access to Project Genie — Google's tool for generating 3D worlds. For the average Czech user, what matters is that Gemini remains available for free in its basic version and that the AI Plus tier (the equivalent of ChatGPT Plus) stays at a more affordable price of around $20 per month.

All listed prices apply to the US market. Google typically converts European prices at a 1:1 rate in euros, so €100 or €200 per month can be expected, respectively.

CodeMender: Google takes on Anthropic's Mythos

Beyond the mainstream consumer announcements, Google also revealed the API opening for CodeMender — an AI code security agent that can not only detect vulnerabilities but also fix them directly. This is a direct competitor to Claude Mythos from Anthropic, which shocked the world in April 2026 with its ability to find unknown security holes in critical systems.

Google CEO Sundar Pichai acknowledged at a press briefing: "What Mythos has accomplished — and credit where it's due — is show that the largest models have enormous value precisely in security scenarios." Google is already in talks with governments and enterprises about deploying CodeMender to audit their systems.

What does this mean for Czech users and businesses?

Gemini 3.5 Flash and the new redesign are available immediately in the Czech Republic as well — just update the Gemini app on Android or iOS, or visit the web version. Czech has long been supported in Gemini, so you can communicate with the new model in your native language.

For businesses, the key point is that Google has significantly improved agentic capabilities and safety features with Gemini 3.5 Flash. This paves the way for deploying AI in production with a lower risk of unwanted outputs — exactly what European companies under the AI Act's oversight need.

Gemini Spark, Omni Flash, and Daily Brief are currently limited to the US market, but Google has a history of gradual expansion into Europe. If you're keen to try out agents that work in the background or video generation from anything, it's worth watching the official Google blog, where European launch dates will appear first.