Genie 4 Access and Capabilities for Builders

Only a limited group of trusted testers can use Genie 4 today. Google announced the new world model on October 27, 2026, but has not opened it to the general public. Google has published no price, model ID, or general release date.

At Layer3Labs, we build custom agents and enterprise integrations for small and mid-sized businesses.

Google says the model is coming soon to Google AI Ultra subscribers in Project Genie. Early access partners in the Gemini API and Google AI Studio represent a separate planned group. You can review the initial research disclosure in the Google DeepMind blog or read the launch overview on the official Google blog. Until Google publishes a model ID and pricing, treat it as a watch-list item rather than an active engineering target.

Who Can Use Genie 4 Today

Genie 4 is a research preview. Google says it has started rolling out the model to a limited group of trusted testers across creative industries, journalism, education, gaming and engineering. Google says Genie 4 has strict safety guardrails consistent with its policies for other models and went through internal safety evaluations before the preview. All generated video and audio includes SynthID watermarking.

If you do not belong to that initial testing group, you cannot run Genie 4 right now.

A waitlist form is linked through Project Genie for users seeking future access, according to the Google DeepMind post.

What You Can Try Today in Project Genie

Outside the trusted-tester group, the only way to try a Google world model hands-on is Project Genie. It runs on Genie 3 alongside Nano Banana Pro and Gemini. Google announced this prototype web app on January 29, 2026, as an experimental environment for creating and navigating real-time interactive worlds.

Access to Project Genie is limited to Google AI Ultra subscribers in the United States who are 18 years of age or older. Google AI Ultra for Business users are excluded from the prototype. Google says access will expand to more territories "in due course." The app is at labs.google/projectgenie.

Google lists these limits for Project Genie. Generation is limited to 60 seconds, and worlds may not look fully realistic or follow prompts, images or real-world physics. Characters can be harder to control and may have higher control latency, and promptable events that change the world during exploration are not yet included.

Within the web application, qualifying subscribers can use three workflows:

• World sketching: Create a world and a character from text and images, preview it with Nano Banana Pro, and choose a first-person or third-person view.
• World exploration: Move through the world as Genie 3 generates the path ahead in real time, and adjust the camera.
• World remixing: Build on existing worlds, browse a curated gallery, use the randomizer, and download videos of your worlds and explorations.

What Google Has Published About Genie 3 and Genie 4

Google announced Genie 3 on August 5, 2025. Google says Genie 3 environments remain largely consistent for several minutes. Among the limits Google lists for Genie 3 are a limited action space for the agent and imperfect geographic accuracy for real-world locations. Google also lists difficulty modeling complex interactions between independent agents, and says legible text often appears only when it is included in the input world description.

Google describes Genie 4 as generating audio natively, and Google's Genie 3 post does not mention audio. Google has published no resolution, frame rate or session duration limits for Genie 4.

What Genie 4 Does

Genie 4 operates as an interactive world model where users navigate environments using standard keyboard controls while directing the simulation through text prompts. It simulates physical scenarios such as aerodynamics and collisions. Google lists complex physical interactions as an open area, so treat the physics as a preview feature.

A creator can type a natural-language prompt to modify the weather, reposition the camera, or insert a person or object while the simulation is running. Characters within the scene move naturally, respond to their surroundings, and accept real-time behavioral guidance from user prompts.

Genie 4 generates audio natively alongside visual frames. Each simulation generates ambient background sounds, such as wind or waves, paired with interactive character dialogue that adapts to inputs and environmental events.

Gemini Omni anchors simulations in real-world appearance and behavior, while other Gemini models handle reasoning for coherence. Gemini 3.8 Flash acts as a guiding agent that plans steps, verifies progress and adapts when the user intervenes.

In one robotics experiment, Google used Gemini 4 Argon and Genie 4 to simulate a humanoid robot playing checkers, with task steps delivered as text captions. Google has published no model ID for Genie 4, so there is nothing to build a similar setup against yet.

The Limits Google States

Google acknowledges several technical limitations in the research preview. Visual details and object permanence can drift, especially in longer sessions or after multiple edits.

Google says Genie 4 keeps environments coherent over short sessions. If you get early access, plan any test around short sessions with few edits.

Google lists five open areas:

• Complex physical interactions
• Real-time audio accuracy
• Legible text rendering
• Spatial accuracy
• Long-term consistency

Five Things Google Has Not Published About Genie 4

Google announced Genie 4 on October 27, 2026, but left five operational details unpublished: pricing, a model ID, regions, technical performance limits, and release dates. The missing specifications block practical decisions on coding, budgeting, and scheduling. Check the Google DeepMind post and the Google post for published updates on these specifications before committing to a development schedule.

• Model ID: Google has not published a model ID for Genie 4, which means you cannot write code against Genie 4.
• Pricing: Google has not published pricing, which leaves you unable to budget or say whether it will be free.
• Regions: Google has not published regions, so you cannot tell whether your location qualifies.
• Resolution, frame rate, and session limits: Google says only that environments stay coherent over short sessions, so you cannot judge whether Genie 4 fits a product.
• Release dates: Google says Project Genie is coming soon for Google AI Ultra subscribers. It gives no date for Gemini API and Google AI Studio partners, so you cannot schedule work.

Who Should Ignore Genie 4 for Now

Software engineers building transactional business software, customer service agents, and automated internal workflows should ignore Genie 4 for now. Because there is no model ID, API access, or price to plan against, teams running business workflows should not plan a build around it.

Google has not published pricing, so no one can say yet whether Genie 4 will be free or paid.

Treat Genie 4 as a watch-list item. This changes if Google publishes a model ID or pricing, or opens the Gemini API or Google AI Studio to general access. Until then, join the Project Genie waitlist if you are a creator or simulation team.