ModelJEVJev

TypeSafe AI launches Jev 1.14.0 with image support

Original title: Jev: For the builders AND dreamers

Plain language

What it is: TypeSafe AI has released an updated version of its AI model, Jev 1.14.0. They also introduced experimental features including image support, longer text context, and a new data type.
How it works: The experimental features can be turned on using special settings in the API. The new model version brings general improvements, with the biggest gains in multi-lingual support.
Where it is available: These updates are available through the company's API. The new version is currently accessible under the preview name and will become the default version in one week.
Example: Users can now include images directly inside the instructions or criteria they send to the Jev model.

What it is: TypeSafe AI released jev-1.14.0 and introduced experimental danger zone features. These features include multimodal image support, longer context lengths, and a new output type called box.
How it works: Developers can opt into these experimental features by passing specific API headers, such as TypeSafe-Danger-Zone-Images and TypeSafe-Danger-Zone-Long-Context. The new box type returns a single bounding box and is exclusive to jev-1.14.0.
Where it is available: The features are available in the TypeSafe AI API. The jev-1.14.0 model is currently under the jev-preview alias and will become jev-latest in a week, while jev-1.13.0 remains active with a 60-day deprecation warning period.
Example: A developer can pass images anywhere within the JSON interface, such as in the state, instructions, or criteria, by enabling the image header.

The headline and explanation on this page are based on TypeSafe AI's post. The full details are in the original.

Jev: For the builders AND dreamers