Google Launches Gemini 3 With Immediate Search Integration
Google releases its most advanced AI model, Gemini 3, integrating it directly into Search on day one. The new model sets record highs across major reasoning and coding benchmarks.
Google unleashes Gemini 3, its most advanced AI model to date, marking its most aggressive product launch since the generative AI wars began. Unlike previous releases that took weeks to roll out, this model goes live immediately across Google Search, the Gemini app, AI Studio, and enterprise products. For the first time in the company's history, Google embeds its latest frontier model directly into Search on launch day, showcasing both technical confidence and competitive urgency in a rapidly moving market.
The new model dominates every major benchmark that matters for real-world applications. Gemini 3 Pro scores 37.4% on Humanity’s Last Exam, crushing the previous record held by GPT-5 Pro, and claims the top spot on LMArena with an Elo rating of 1501. Additional impressive scores include 91.9% on GPQA Diamond for graduate-level scientific reasoning, 87.6% on Video-MMMU for video comprehension, and 1487 Elo on WebDev Arena for coding capabilities.
For users who need even more computational power, Google introduces Gemini 3 Deep Think, a research-intensive variant optimized for the hardest reasoning tasks. This enhanced version pushes the limits even further, achieving 41.0% on Humanity’s Last Exam without tools and 93.8% on GPQA Diamond. This launch arrives just seven months after Gemini 2.5 and less than a week after OpenAI's GPT-5.1 release, highlighting the blistering pace of frontier AI development.