What matters in AI.

Subscribe

Google opens its Genie 4 world model to developers through its API

Each Genie 4 session ends after 120 seconds, and video costs about $0.10 for each second.

Google opened Genie 4, a world model from Google DeepMind, to developers in the Gemini API and Google AI Studio. A world model makes a scene that the user controls as it plays. It came one day after Genie 4 came to Google AI Ultra subscribers. Google says it is its first world model for developers, and it is not free to use.

How it works

A developer gives text, an image or the two. Genie 4 makes a continuous world with sound. The user moves with the keyboard and types new prompts to change the weather, light or objects. Google says each output has a SynthID mark.

The numbers

  • Video costs about $0.10 for each second.
  • A session ends after 120 seconds.
  • Keyboard moves show after about half a second.
  • Text prompts show after one to two seconds.

What is not known

Google gave no test results for the quality of the worlds or for robot training in them. Google gives prototypes and robot training as uses. Google also says world simulation is a step to AGI (AI as good as a person at all tasks).

Insight: Google adds more tools to one platform for developers

Genie 4 uses other Gemini models for its look and its logic. It comes with Gemini 4 Hydrogen, a coding model, and new Gemini assistants for cloud work, all in November. This reads as a move to make the Gemini API the one source of developer tools, and each tool pulls developers to the other tools. The next move to monitor is the price for each second.

Sources

  1. Google opens Genie 4 world model to developers in Gemini APItestingcatalog.com
AI MATTER · NEWS · AI MATTER · NEWS ·19 NOV2026

Posted

Tags

Learn the terms in this story

More in Models

All Models news