• by Laurie Sullivan , Yesterday

Google Gemini 3 Pro, which launched Tuesday, brings a new level of depth and nuance to interactions.

The AI models avoid flattery and aim to provide genuine insight. The goal is for the models to act as a "true thought partner" rather than offering responses that flatter or provide clichés, the Google DeepMind team explained in a post, adding that the models tell "you what you need to hear, not just what you want to hear." 

The latest Google Gemini 3 AI models mark a significant change because they introduce advanced reasoning and multimodal understanding for a range of tasks -- from learning and research to complex coding and software development.

Google Chief Executive Officer Sundar Pichai introduced the models in a video Tuesday, describing them as the most capable models to date because they support inputs and outputs across media including text, images, video, audio and code.

The models demonstrate advanced performance in reasoning and planning, with the ability to see, hear and understand context, enabling them to complete complex challenges in math and human behavior. 

They have the ability to understand context, human behavior, and nuance that allow for segmented personalization, which can help advertisers to move beyond basic demographics and increase engagement and conversion rates.

While personalization exists, AI targeting uses sophisticated machine learning to understand complex context, predict future behavior in real-time, and tailor messages to the individual consumer in a "segment of one" targeting model that is only possible through these advancements.

Benchmark results show Gemini 3 Pro outperforming its predecessor, Gemini 2.5 Pro, across key evaluation metrics with a leading score of 1501 Elo on the LMArena Leaderboard and scoring the highest marks on two exams that assess reasoning, including a 37.5% result without tools on Humanity's Last Exam and a 91.9% on the GPQA Diamond.