G
GOOGLE / Genie 4

Genie 4

A world model that generates explorable scenes with native video and audio. It supports text instructions, visual references and interactive navigation. Google opened research access to authorized testers on October 27.

01 / SUMMARY

The essential

ORIENTATIONInteractive world simulation

Primary use stated or inferred cautiously from official documentation.

CONTEXTInteractive scenes and controls; no published token limit

Maximum input capacity when published by the source.

EXITReal-time video and audio; sessions lasting a few minutes

Documented maximum generation limit.

ACCESSAuthorized testers only; public waitlist

Verified availability channels.

MODEL TYPEVideo · Voice and audio · Multimodal

Primary functional family and verified specialties.

LICENSEProprietary

Declared terms for API access, model weights, or self-hosting.

02 / TECHNICAL SHEET

Limits and integration

API IDGenie 4
Model typeVideo · Voice and audio · Multimodal
Access modelProprietary · restricted research
LicenseProprietary
DeploymentHosted research service
Release27 OCT 2026
Knowledge cutoffNot published
EntranceText · Image · Navigation controls
ExitVideo · Audio
Context windowInteractive scenes and controls; no published token limit
maximum outputReal-time video and audio; sessions lasting a few minutes
ReasoningNot published
Published tools
Structured OutingsNot published
Batch processingNot published
Prompt cacheNot published
Fine-tuningNot published
Verified platformsGoogle DeepMind
BEST SUITED FOR

Creative previsualization, experience prototypes and exploration of scenarios for agents, within authorized access.

WORTH MONITORING

Project Genie for Ultra subscribers, the Gemini API and AI Studio are still planned for later. No public API identifier has been verified. Object permanence, physics, text and audio or spatial accuracy can fail.

03 / CAPABILITIES

What it can do

01

Orientation

Creative previsualization, experience prototypes and exploration of scenarios for agents, within authorized access.

02

Context

Interactive scenes and controls; no published token limit · Real-time video and audio; sessions lasting a few minutes

03

Tools and integration

Not published

04

Access

Authorized testers only; public waitlist

Input modalities
TextImageNavigation controls
04 / PRICES

Documented cost

ConceptWorthUnit/condition
Standard inputNo verified public pricingStandard API rate
Cached inputNot publishedReading reused prefixes
Cache write or storageNot publishedThe condition varies by provider
Standard outputNo verified public pricingMay include reasoning tokens
Batch inputNot publishedAsynchronous processing
Batch outputNot publishedAsynchronous processing

No prices, quotas or downloadable weights have been published. An Ultra subscription does not establish current access to Genie 4. Preview terms depend on the access program.

i Prices change and may depend on level, region or context length. Check the source before making a decision.

05 / EVALUATIONS

How to read the results

“

Provider demonstrations are not a reproducible benchmark. We assign no scores and carry over no resolution, frame rate or limits from earlier versions. Simulation does not establish transfer to physical robots.

i No verifiable quantitative benchmarks published for this version.

Inferama only highlights a “best result” when the metric, test set, configuration, and date allow for an equivalent comparison. The supplier's figures are presented as claims from its own source.

06 / VERSIONS

Chronology

Genie 4

Preview for authorized testers; video and audio marked with SynthID. Public access is still pending.

ANALYSIS

Related analysis

07 / SOURCES

Traceability