Gemini 4 Carbon Available to All, Prices Doubling on December 1

Gemini 4 Carbon is now available to all users, with a launch price of $2 per million inputs and $10 per million outputs. An increase to $4 and $20 is scheduled for December 1. Online demonstrations and tests with the Antigravity IDE showcase the model's ability to coordinate sub-agents to generate complex 3D projects.

Demonstrations Show Multi-Agent Orchestration for a 3D Game

A YouTube video highlights the use of the /swarm feature of Antigravity with Gemini 4 Carbon. In this example, the system assembles an open-world 3D game where players explore the Google campus in Mountain View as a pelican riding a GBike. The coordination of sub-agents allows for modeling assets in Blender from named references Omni and Nano Banana, building the browser game in Three.js with WebGL shaders, and adding procedural sound effects via Web Audio accompanied by Lyria music.

Widespread Access and Launch Pricing, with Announced Increase

Gemini 4 Carbon is now accessible to everyone. At launch, the pricing is $2 per million inputs and $10 per million outputs, amounts equivalent to those of GPT-6.1 Sol and Claude Sonnet 5.5. Prices are expected to rise to $4 and $20 on December 1, a level announced to be identical to that of Opus 5.5.

Levels of "Reflection" and Testing via Antigravity

The model offers three levels of "reflection" difficulty—low, medium, and high—with a default setting on medium. A high-level trial produced a pelican-like result. Tests via Google’s Antigravity IDE revisited a prompt initially used with Claude Fable 5.2: generating a three.js animation where a mech-style pelican stomps through Manhattan, Queens, and Brooklyn using OpenStreetMap data. The output was provided in HTML, then converted into a standalone file without external dependencies before being captured on video. The LLM-Gemini 0.36 environment serves as a plugin to access the Gemini family. Gemini 4 Carbon had been announced two weeks earlier for "a subset of cyber defense advocates" and is presented as Google's best Gemini model.