Google says Gemini 4 Carbon is now globally available for developers and enterprises through the API in AI Studio and Gemini platform. Google AI Ultra and Pro subscribers can also use it in Google Antigravity, the Gemini app, and AI Mode in Search. The rollout puts one model across build tools and consumer products at once.
AI Studio and Gemini platform
Google says Gemini 4 Carbon was built to solve hard problems. It can reason, code, and coordinate tools, which is the practical split that matters for teams trying to move from prompts to working output.
The company also says the model excels in deep comprehension and long-horizon workflows. That points to work that spans many steps, not one-off answers, and it is the main reason developers will care about the wider API access.
Google Antigravity and the app
Earlier demonstrations show what Google wants readers to picture. Gemini 4 Carbon wrote 5,000 lines of math to render a 3D sunlit library from scratch, and Googlers used it to prototype complex 3D visualizations, high-fidelity games, and interactive animations.
It also generated a custom audiovisual tool that traces musical scale lines from frame to frame when prompted to build an app that overlays shapes across videos. In Google Antigravity, it orchestrated subagents, Blender MCP, and Nano Banana to turn pictures of a real desk into a playable 3D track.
NASA files and PDFs
Google says Gemini 4 Carbon parsed more than 80,000 raw files from NASA and turned them into an executive-ready deck. The files included weather logs and engineering, health, and science data accumulated across 72 flights on Mars, which suggests the model is being pitched for large document sets as much as for code.
It also used Nano Banana and Veo to transform scientific concepts in a PDF into an illustrated animation in the Gemini app. Google says it generated a self-navigating 3D periodic table with animated atoms and frame-accurate audio timing to the lyrics of The Elements by Tom Lehrer.
Google AI Ultra and Pro
The new access pattern matters because the same model now reaches developers, enterprises, and subscribers through different Google surfaces. That broad spread is useful for people testing the model in code, in the app, or inside Search without switching systems.
The missing piece is performance comparison. Google describes Gemini 4 Carbon as smarter and more capable, but the examples do not provide independent benchmarks or a clean before-and-after measure against older models or rivals.
