Google Launches Gemini 3: Unprecedented Features Rivaling the Most Powerful AI Models
Google officially announced today the launch of Gemini 3, which it described as its most powerful model to date and the world's best model in multimodal understanding, in addition to a significant leap in AI programming and AI-agent capabilities.
Gemini 3 represents a major step forward in the development of AI models, thanks to its deep reasoning capabilities, richer visualizations, and improved performance across various cognitive and creative tasks, making it a comprehensive model approaching expert level in many fields.
Gemini 3 Pro: Unprecedented Reasoning Accuracy
Google confirmed that Gemini 3 Pro is now available across a wide range of the company's products, aiming to provide users with a smarter everyday experience, whether for learning, planning, or creating.
The new model delivers superior performance compared to its predecessor, version 2.5 Pro, across all key benchmarks. It achieved world-class scores, most notably LMArena's top score of 1501 Elo, a PhD-level reasoning ability, and a score of 37.5% on the Humanity’s Last Exam without tools. It also demonstrated exceptional performance on the GPQA Diamond test with a score of 91.9%.
Furthermore, it achieved a new world benchmark in MathArena Apex with a score of 23.4%, along with multi-mode excellence in MMMU-Pro (81%) and Video-MMMU (87.6%), and leading-edge information accuracy in SimpleQA Verified (72.1%).
These results confirm the model's ability to handle complex topics in science and mathematics and solve problems with high reliability.
Gemini 3 Pro delivers a new level of depth in its answers, relying on direct clarity and avoiding embellishment. It acts as a true thinking partner, providing users with new ways to understand and formulate information, whether through generating high-resolution visualizations and graphics, translating complex scientific concepts, engaging in creative brainstorming, or analyzing videos.
Alongside this new model, Google launched the Google Antigravity development platform, which completely reimagines the developer experience. This platform enables the use of Gemini 3 as intelligent agents capable of coding, executing complex tasks, interacting with the editor, terminal, and browser, validating code, and managing tasks simultaneously.