Google Unveils Gemini 3: The Ultimate AI Innovation

Google has launched Gemini 3, a new series of AI models that the company claims are its most intelligent and factually accurate yet. These models aim to put Google ahead of OpenAI after the rocky launch of GPT-5, positioning Google at the forefront of consumer-focused AI models.

Gemini 3 Pro, Google’s new flagship AI model, is now accessible to everyone in the Gemini app on the first day of its launch. It is also available to subscribers within Search. Tulsee Doshi, Google DeepMind’s senior director and head of product, believes that this new model will bring the company closer to making information universally accessible and useful as the search engine continues to evolve.

Google is stepping out of the paradigm of just text responses with Gemini 3 Pro, offering a much richer and more complete view of information. The model is natively multimodal, capable of processing text, images, and audio simultaneously, rather than separately. For example, Gemini 3 Pro could translate photos of recipes into a cookbook or create interactive flashcards based on video lectures.

Improvements from Gemini 3 Pro can be seen across Google’s suite of products, including the Gemini app, where users can build more full-featured programs inside the built-in workspace, Canvas. The upgraded AI model also enables generative interfaces, allowing Gemini 3 Pro to create visual, magazine-style formats with browseable pictures or dynamic layouts with custom user interfaces.

In AI Mode, Gemini 3 Pro will present users with visual elements like images, tables, grids, and simulations based on their queries. It can perform more searches using an upgraded version of Google’s query fan-out technique, which not only breaks down questions into searchable bits but also better understands intent to find new content.

Google emphasizes that Gemini 3 Pro offers smart, concise, and direct responses, trading cliche and flattery for genuine insights. The model shows reduced sycophancy compared to previous AI models, addressing issues faced by OpenAI with ChatGPT earlier this year.

Gemini 3 Pro comes with enhanced reasoning and agentic capabilities, allowing it to complete more complex tasks and plan ahead over longer horizons. An experimental Gemini Agent feature powered by the AI model can perform tasks on behalf of users within the Gemini app, such as organizing emails or booking travel.

Gemini 3 Pro has secured the top spot on LMArena’s leaderboard, a platform for benchmarking AI models. A Deep Think mode further enhances the model’s reasoning capabilities, though it is currently only available to safety testers.

Gemini 3 Pro is now available in the Gemini app for everyone, while Google AI Pro and Ultra subscribers in the US can access the model in AI Mode by selecting “Thinking” from the model dropdown. Gemini Agent will first roll out to AI Ultra subscribers.