Google released Gemini 3 yesterday with immediate availability across its entire ecosystem. The company says the model sets new performance benchmarks: Gemini 3 scored 37.4 on Humanity’s Last Exam, the highest on record and substantially above GPT-5 Pro’s 31.64. It topped the LMArena leaderboard for user satisfaction and introduced generative UI capabilities that create interactive interfaces, not just text responses.
Google also launched Antigravity, an agentic development platform that lets developers work at task level rather than code level, as well as Gemini 3 Deep Think, which is an enhanced reasoning mode coming soon to Ultra subscribers after additional safety testing.
Gemini 3 arrives seven months after Gemini 2.5, less than a week after OpenAI’s GPT 5.1 update, and two months after Anthropic’s Sonnet 4.5. The pace of frontier model releases has compressed to weeks and months, rather than years. Google skipped the cautious rollout strategy it used for Gemini 1.0, when image generation errors and “woke” controversies forced the company to pull back.
Gemini 3 can create custom layouts with interactive elements, comparison tools, and dynamic visualizations tailored to each query. A question about physics generates an interactive simulation. A request for financial analysis produces a working loan calculator. These AI-generated experiences are way more fun than traditional results; if they catch on, we could see new economic models for content discovery.
Author’s note: This is not a sponsored post. I am the author of this article and it expresses my own opinions. I am not, nor is my company, receiving compensation for it. This work was created with the assistance of various generative AI models.