TypeSafe AI, the creators of Jev, announced the launch of Jev 1.14 under the alias ‘Jev Preview’, an upgraded version of their flagship AI model, that now allows users to upload images in their queries and introduces ‘Box’ System One QuestionType and longer context upto to 262K tokens.
Unlike conventional large language models that generate free-form text, Jev is a System One model that returns typed, calibrated decisions with associated confidence scores, making it suited for direct integration into software workflows such as routing, classification, and structured judgment tasks.
On Wednesday, TypeSafe AI had a limited preview launch of Jev 1.14 on their Discord channel late Wednesday with over 500 virtual attendees.
we announced Jev 1.14 on 10.14 at 1:14pm PT 🔥
a new and improved model, along with experimental features for our dreamers:
• images
• a new box type
• longer contextsurprise, Jev was already able to see you! 👁️
BUT our bar for production-grade intelligence is quite high, so builders beware, these are preview features.
enter the ~*~ danger zone ~*~ at your own risk. 🚨
— TypeSafe AI (@typesafeai) October 15, 2026
The AI company deliberately chose a conservative approach, calling the image option and longer context as ‘danger zone’ for builders and developers and have referred to the upgrades as “less ready” for automation and use-cases.
“See these new capabilities as a continuation of what we’ve built over 2 years of stealth. They’re less ready for automation, but very fun to play with. They’re also an early glimpse into what we’re focused on making reliable enough to meet TypeSafe standards,” read a statement from latest blogpost by TypeSafe AI.
Under the new upgrades, TypeSafe AI has now enabled multimodality for both Jev 1.14 and flagship Jev model The new model Jev 1.14 and users can now send images in a system One request via in the state, inside a question’s instructions and as the criteria a question is judged against.
Similarly, TypeSafe AI has introduced Box, a System One question type for users to help them find something in any image while processing request to the System One model. The company has warned users that their new upgrades of images and box questions are not yet “upto their standards” and encouraged developers to use them to build only prototypes.
The new version, available via the jev-preview alias, delivers subtle gains across the board, with the most notable improvements expected in multilingual performance. Pricing remains unchanged at $42 per billion input tokens, with output tokens free. Core context limits stay at 64k tokens per request and 32k per question by default, though an experimental long-context mode expands both to 262k tokens when the TypeSafe-Danger-Zone-Long-Context header is enabled. TypeSafe cautions that longer contexts reduce reliability.
Also Read: TypeSafe AI Pauses Jev Signups After ‘Immense Swell of Demand’
