Google Deep MindOctober 27, 2026

Genie 4

Google DeepMind’s interactive world model generates synchronized audio and video at up to 720p and 24 fps.

Text + Image inputAudio + Video outputClosed / API only

At a glance

Context window
—
Max output
—
Text input / output · per 1M tokens
$1.50 / $9.00
Knowledge cutoff
—

API pricing

Text rates in USD per million tokens; other billing units shown below · effective 2026-11-19

Input
$1.50
Output
$9.00

Blended price4:1 input:output · same basis as the price ranking

$3.00/ 1M tokens

Image pricing

  • image input$1.50 / 1M tokens

    Reference image input, same rate as text input.

Estimate running cost for Genie 4

Specifications

Provider
Google Deep Mind
Model type
Multimodal model
Released
October 27, 2026
Architecture
Undisclosed
Input
Text, Image
Output
Audio, Video
Weights
Closed / API only
Commercial use
Proprietary

Model analysis

Genie 4 is Google DeepMind’s real-time interactive world model, announced on October 27, 2026. It produces video at up to 720p and 24 frames per second with synchronized audio. Its interactive demonstrations are compelling, but reproducible quantitative benchmark results have not been published.

Live steering comes with limits on long-term consistency

The model takes text, reference images and real-time keyboard and mouse actions. Users can change weather, objects or characters during a session while the system generates visuals, ambient sound, music and dialogue. Google describes a Transformer-based system adapted from Gemini Omni Flash, with Gemini providing reasoning and real-time steering. Its audio and video carry SynthID watermarks. Google also says that visual details and object permanence can drift over longer sessions or after repeated edits; complex physics, spatial accuracy and text rendering remain limitations.

Video token pricing includes synchronized audio

Available as genie-4 on the paid Gemini API tier, standard pricing is $1.50 per million input text or reference-image tokens, $9 per million output text tokens, and $17.50 per million video output tokens with audio included. At 24 fps, 5,760 video tokens per second imply about $0.1008 per second. Costs accumulate during generation; Google says to end the session to stop billing. Free tier, Batch, Flex and Priority are unavailable.

Public demos do not establish reproducible physics performance

The model card reports human and automated evaluations and qualitative improvements over Genie 3, but provides no scores, sample sizes or full evaluation protocols. Parameter count, context window and a complete supported-language list are undisclosed. Genie 4 can support creative previsualization, interactive prototypes and simulation experiments; physical accuracy and transfer to real-world robotics remain unproven by the public evidence. Project Genie offers Genie 4 to eligible Google AI Ultra subscribers. Longer sessions and panorama seeding are described for Genie 3 and should not be assumed to apply to Genie 4.

Sources: Google DeepMind announcement, model card, API pricing, and Project Genie update.

Last reviewed 2026-11-20

Community discussion

Share the API, setup, and tradeoffs you actually tested. Rate the model once; comment as often as you like.

Used this model? Sign in with Google to rate it and share what worked, what did not, and which API or setup you tested.

No community comments yet.