Google's Gemini 3 rises to the top

Alongside a new agent-first development platform called Antigravity, Google unveiled Gemini 3, a new SOTA model that it claims ushers in a "new era of intelligence." Gemini 3 topped a number of leaderboards and benchmarks. The specifics: Gemini 3 and Deep Think surpass GPT-5's prior scores on reasoning standards like Humanity's Last Exam and ARC-AGI-2. The model only trails Claude Sonnet 4.5 in coding, but it leads in scientific understanding, math, multimodal reasoning, and tool utilization. Gemini 3's capacity to create generative user interfaces and visual layouts on the fly is demonstrated by its interaction with AI Mode in Search. Additionally, Google introduced Antigravity, a free agentic coding platform featuring multi-agent orchestration, asynchronous processes, and browser control. For the first time in a long time, Google has overtaken OpenAI with the eagerly anticipated Gemini 3. Your move, Sama. As the business moves into the next generation of frontier models, the internet behemoth is undoubtedly displaying some real power thanks to its limitless resources and integrations throughout its vast product ecosystem.