Summary

Iris-3B is a pixel-space generative model designed to serve as both an image generator and a general visual learner. It works directly in pixel space rather than compressing images through a VAE, exposing generative representations for visual tasks.

Key Features

Pixel-space image generation.
No VAE bottleneck.
General visual representations.
Dense vision-task experiments.
Released scaling recipes.
Public model weights.
A standard demo.
A Turbo demonstration.

About Iris-3B


Developers can explore detailed generation and dense visual understanding using released weights, scaling recipes, and demonstration interfaces. Core capabilities include Pixel-space image generation, No VAE bottleneck, General visual representations, and Dense vision-task experiments.


Public code or model weights are linked from the project, with setup instructions for running or adapting the system. Additional characteristics include Released scaling recipes, Public model weights, A standard demo, and A Turbo demonstration.

Get more likes & reach the top of search results by adding this button on your site!

Embed button preview - Light theme
Embed button preview - Dark theme
TurboType Banner

Subscribe to the AI Search Newsletter

Get top updates in AI to your inbox every weekend. It's free!