Article
CommunityJev Media & Talks
Jev 1.14 Release, Reported by MarkTechPost
MarkTechPost's write-up of the Jev 1.14 release: images, a box question type and 262k long context as experimental danger-zone features, unchanged $0.042 per million input token pricing, and TypeSafe's own note that none are production-ready.
- Format
- Article
- Reading time
- 3 min
- Published by
- Community
- Author
- Asif Razzaq
- Last verified
- Oct 2026
Used inJev for Software & DevTools
Highlights
- Jev 1.14 ships as jev-preview with images, a box question type and 262k context, all behind opt-in danger-zone headers.
- Image limits: PNG or JPEG stills up to 16 MiB per file, 256 to 1568 pixels per side and a 32 MiB body, billed as input tokens.
- Box takes one bounding box per question and always returns one; pair it with a Noul that checks whether the target exists.
- In TypeSafe's test, a 4-question edge check caught 30 of 36 cut-off crops; one whole-image question caught 26, an outline question 20.
- Pricing is unchanged: $42 per billion or $0.042 per million input tokens, output free; jev-latest moves to 1.14.0 in a week.
Watch out
A press summary of TypeSafe's announcement, not an independent test. The box-check figures and the intelligence comparison are TypeSafe's own, its charts compare Jev only against open-weight models, and Jev 1.14 is a hosted model with no open weights, no disclosed parameters and no public benchmark scores cited.
Questions about this entry
Short answers drawn from the linked project or page.
Images, a box question type that answers with where something is in an image, and long context: an opt-in header raises the defaults of 64k tokens per request and 32k per question to 262k. TypeSafe says none of the three are ready for production.
Each image is listed once in a top-level assets map under a reference string and can be placed anywhere in state, instructions or criteria; the API swaps each reference for the image. Jev 1.13 and Jev 1.14 both accept images behind the images header.
No. Box always returns a box, so a missing target still produces four edges and a confidence; TypeSafe tells callers to pair it with a Noul that checks whether the target exists at all. Box takes no criteria.
Four edge-check questions caught 30 of 36 cut-off crops, a single question asking whether all of it is inside caught 26, and drawing the box as an outline and asking about it caught 20. These are TypeSafe's numbers, with no public scores.
PNG or JPEG still images only, up to 16 MiB per file, with each side between 256 and 1568 pixels and a 32 MiB request body cap. Images bill as input tokens at the normal price.
It warns that all models perform worse on images than on text and recommends the simplest possible questions, and it says longer context reduces reliability. That is why long context is opt-in and why the dev team still recommends short contexts.
The rollout moves jev-latest to 1.14.0 about a week after launch. jev-1.13.0 stays available with 60 days of notice before formal deprecation, and jev-1.14.0 itself gets 30 days of notice before its own deprecation.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: A doorbell camera on Jev 1.14
More Jev resources like this
Tech & Business newswire item on TypeSafe opening experimental API access to image inputs and longer context for Jev, alongside the Jev 1.14.0 release through jev-preview and a new box output type.
Jev Media & TalksArticle#community#news#images
Community
GIGAZINE's report on Liquid AI's d1, which it describes as the first decision model to beat Jev on Hugging Face's Decision Index, and a short survey of the clone wave that followed Jev's September launch.
Jev Media & TalksArticle#community#news#benchmarks
Community
RSS.bz's compilation of TypeSafe's launch posts for Jev 1.14, released October 14 at 1:14pm PT with experimental preview features: image support, a new box type and longer context, plus internal benchmark guesses.
Jev Media & TalksArticle#community#news#model-release
Community
Langfuse's September 18 post presents Jev as a decision model for routing, classification and eval verdicts, and walks through scoring production traces with Jev-as-a-judge in Langfuse.
Jev Media & TalksArticle#community#press#explainer
Community
A community roundup of every public Jev evaluation: TypeSafe's own workflow evals plus at least twenty independent builder benchmarks. The site reports speed medians of 127-422 ms per call leading every LLM compared, while accuracy splits by task.
Jev Media & TalksArticle#community#press#calibration
Community
A playground operator's audit of every published Jev number: TypeSafe's four-workflow evals, the favourable homepage demo, independent tests and its own calls. It concludes the price and speed are real and the intelligence claim is roughly mid-tier frontier on decision workflows.
Jev Media & TalksArticle#community#press#explainer
Community
From the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.