TypeSafe AI has released Jev 1.14, introducing image input capabilities and a bounding box data type for object localization alongside a fourfold expansion in context capacity. The update enables multimodal functionality across existing Jev models, allowing images to be combined with text inputs. The new version delivers broad performance improvements, with multilingual support seeing the most significant gains. The jev-preview endpoint now points to jev-1.14.0, though TypeSafe AI notes that these features remain experimental previews and have not yet reached production-grade standards.