Ideogram AI just released its latest text-to-image generation model, and this time it comes with a twist that matters: you can download the whole thing and run it yourself. Ideogram 4.0, which launched on June 3, represents the company’s most ambitious bet yet on openness as a competitive strategy in the increasingly crowded generative media landscape.

The model weights are available on GitHub under the repository ideogram-oss/ideogram4. That means developers, researchers, and tinkerers can run local inference, fine-tune the model for their own use cases, and build on top of it without asking permission.

What’s actually improved

User feedback has pointed to significant advancements in photorealism and complex typography handling compared to version 3.0, the model’s predecessor.

The model also supports native 2K output. Most open image models generate at lower resolutions and require separate upscaling pipelines to produce print-quality or high-resolution digital assets. Baking 2K output directly into the model simplifies workflows for designers and content creators who need production-ready images without extra steps.

Layout control and prompt fidelity are also reportedly improved.

The company behind the model

Ideogram AI was founded in 2022 by a team of former Google Brain researchers, including notable researcher Mohammad Norouzi. The company’s first model was released publicly in August 2023.

After releasing its first model, Ideogram secured an $80 million funding round in early 2024. That capital fueled rapid iteration through versions 1.0, 2.0, 3.0, and now 4.0, each building on the last with particular emphasis on typography and design utility.

Ideogram 4.0 is integrated across all of the company’s plans and its API, meaning both casual users on the platform and developers building applications can access the same model.