Jev 1.14 adds experimental images, bounding boxes and longer context
TypeSafe AI has released Jev 1.14.0 in preview, with small improvements across the model and the largest expected gain in multilingual support. It also introduced experimental image support, longer context and a new bounding-box output type, with warnings against production use.
What changed
Jev 1.14.0 is available under `jev-preview`, and the company says `jev-latest` will switch to it in a week. Experimental image support works through the JSON interface, including on Jev 1.13.0, while the new `box` type returns a single bounding box and requires Jev 1.14.0. Developers can also enable longer context through an API header.
Why it may matter for your work
Developers could test workflows that combine images with structured decisions or need a bounding-box output. The multilingual improvements may help applications serving users in different languages, but the experimental features may need substantial testing before use in everyday operations.
What it can actually do
Accept images in JSON state, instructions or criteria; return a single bounding box using the new `box` type; and process longer context when enabled.
Limits to check
TypeSafe does not vouch for the experimental features for production use. It says image inputs and longer contexts reduce reliability; no context limit or price is stated. The `box` type is available only on Jev 1.14.0. Jev 1.13.0 remains available for now, with intended deprecation warnings of 60 days for 1.13.0 and 30 days for 1.14.0.
“Image support is fully composable via a find-and-replace in `assets`.”From the linked source page