TypeSafe AI's Jev 1.14 is now available on AI Gateway
Source
vercel.com
Author
Nick Oates
Date
Why it matters
Jev returns typed Choice, Score and Boolean answers with probabilities in one parallel request, useful for routing and verification steps. Version 1.14 adds image input and long context.
Key takeaways · AI-distilled
The model is addressed as typesafe-ai/jev-1.14.0 and called through the AI SDK's experimental decide function, where each question declares a type such as boolean plus plain-language instructions.
Images go into the state array as a file part with a media type, alongside text parts, so one request can mix a photo and a description before Jev answers the questions.
Long-context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition → mode is experimental and switched on per request with providerOptions.typesafe.longContext set to true, aimed at documents, logs or conversation histories too long for a standard request.
Vercel says Zero Data Retention and No Training can be enabled per request, and decision calls show up in Gateway logs and custom reporting and count toward budgets like other model calls.
Terms in this piece · Glossary
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
