Google I/O 2026: Google reveals Gemini Omni, Gemini 3.5 Flash with faster AI performance

Google DeepMind CEO Demis Hassabis has officially unveiled Gemini Omni model at I/O 2026. Hassabis says the new model is designed to understand and simulate the physical world and called it a significant leap toward artificial general intelligence (AGI)

During the keynote at I/O 2026, Hassabis explained that Omni combines reasoning capabilities from Google’s other generative media models including Veo, Nano Banana, and Genie.

How is Gemini Omni different?

According to Hassabis, Gemini Omni represents a step change in how artificial intelligence understands physical concepts like kinetic energy and gravity. By functioning as a “world model,” Omni can translate complex ideas into highly accurate visual representations.

During the presentation, Hassabis demonstrated the model’s capabilities by prompting it to “Make a claymation explainer of protein folding.” The AI successfully generated a realistic, scientifically accurate educational video in the requested stop-motion style.

Meanwhile, Gemini Omni can also be used to edit video using natural language. Google says users can upload their own media such a selfie video and prompt the new AI model to change the reality of the scene. Hassabis showcased examples where a user could easily adjust details, alter the artistic style, or add entirely new elements to the environment, such as turning a simple circle into a black hole or bringing a static evening stroll to life.

“Anything becomes a canvas for creating entirely new realities,” Hassabis stated.

“This was always our goal with Gemini, and why we built it to be multimodal from the very start. It was a harder path, but the foundation is now paying off,” Hassabis explained.

The first model in the new family, Gemini Omni Flash, is officially launching today and will be available across Google’s suite of products

Gemini 3.5-Flash launched:

Google has also launched the Gemini 3.5-Flash AI model, which will be the new default model inside the Gemini app. The new model promises significant leaps in coding, real-world workflows, and raw generation speed.

According to Pichai, Gemini 3.5 Flash outperforms the previous generation’s Gemini 3.1 Pro across almost all benchmarks.

The company highlighted a massive improvement in the model’s coding capabilities, as well as an “extraordinary jump” in GDP Val, a specific benchmark designed to measure an AI’s ability to complete real-world, economically valuable tasks. Pichai noted that the industry is moving “beyond AI tools that help us write to agents that help us act.”

“When you look at the intelligence versus output speed, it’s in a whole league of its own in the top right quadrant when looking at output tokens per second,” Pichai said during the presentation. “It’s four times faster than other frontier models, and it’s an incredible delight to use.”

Gemini 3.5 Flash already being used by Google:

Google says Gemini 3.5 Flash has already dramatically accelerated software development at Google through “Antigravity 2.0” the company’s reimagined agent-first development platform.

Google claims that the integration of the new model has created a massive spike in internal AI usage.