AI news · October 16, 2026
TypeSafe ships Jev 1.14 with experimental image support
TypeSafe shipped Jev 1.14 with experimental image input and a new box type, while Jev 1.13 remains available, so existing agents can test visual tasks without a forced migration.
- current release
- 1.14.0
- planned deprecation warning
- 60 days
TypeSafe announced Jev 1.14 on October 14 with experimental image support in its danger-zone mode. The release adds a box bounding-box type, a longer context header and the jev-preview alias; TypeSafe says jev-latest will move to 1.14 after a short period. Jev 1.13 remains available, and the release note gives a planned 60-day warning for its deprecation. TypeSafe also warns that multimodal and long-context behavior is less reliable and that its internal evaluations are not independent.
The useful part is a small visual extension to an existing agent format, not proof of a new general model. The explicit preview and old-version path lower migration risk, while the warning makes image tasks a research feature. Use a pinned version, keep visual actions reviewable and compare the result with the text-only path before moving a customer workflow.
What you can do with it
Try Jev 1.14 in the TypeSafe console with synthetic screenshots and a pinned version. Ask the agent to identify one region and return a box, then check coordinates and failure cases by hand. Keep 1.13 available until the image path passes your own task set.
Our take
This is a useful incremental release for Jev users who need visual input. The dangerous mode label and reliability warning are honest signals, not reasons to ignore it. The feature earns a place only after coordinate accuracy and task completion beat the existing text route.
Links Try itRelease notes
Source: TypeSafe — Made With Models writes the brief; the reporting is theirs.