Overview
- Google reports a verified 77.1% score on the ARC-AGI-2 reasoning benchmark—more than double Gemini 3 Pro—with additional gains on tests like Humanity’s Last Exam.
- The preview is available now for developers through the Gemini API in AI Studio, Antigravity, the Gemini CLI, and Android Studio, for enterprises in Vertex AI and Gemini Enterprise, and for consumers in the Gemini app and NotebookLM.
- Access is tiered, with higher usage limits in the Gemini app and NotebookLM access limited to Google AI Pro and Ultra subscribers.
- Google says this release extends the upgraded core intelligence behind last week’s Deep Think update and lets the company validate multi-step, agentic workflows before general availability.
- Demonstrations highlight practical use cases such as generating animated SVGs from text and building a live ISS telemetry dashboard, while independent leaderboards still show rivals ahead on some coding and text tasks.