🔍 Read the full analysis: I Tried Jev’s New Vision On Images From My Own Life on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
TypeSafe released Jev 1.14 on October 14, adding opt-in image inputs and a bounding-box answer type. Thorsten Meyer’s test of 30 images from his own site found strong product-image matching, but the supplied report cuts off before describing the news-image results. TypeSafe says the new capabilities are not ready for production use without further evaluation.
TypeSafe released Jev 1.14 on October 14, adding image inputs and a bounding-box response type to its decision model, with both capabilities gated behind explicit opt-in headers. In a test published the next day, Thorsten Meyer of ThorstenMeyerAI.com used the image feature to compare 30 of his site’s featured images with headlines, reporting perfect separation in the product-image comparisons but only partial results in the material available here.
Jev is designed to answer structured questions rather than generate prose: users provide a state and typed questions, and the model returns items such as a yes-or-no probability, a choice from a list, or a score with confidence. TypeSafe said image inputs work with jev-1.13.0 and jev-1.14.0. The new box type is available only in 1.14 and returns one bounding box for a specified target, along with confidence.
The image capability is enabled with the request header TypeSafe-Danger-Zone-Images: true. Images can be included in the state, question instructions, or criteria for choices and score levels. TypeSafe’s request format uses an assets map and a text reference to place each image where needed. The company’s documentation sets limits of PNG or JPEG still images, 256 to 1,568 pixels per side, 16 MiB per file and 32 MiB per request; one invalid image can cause the full request to fail.
Meyer tested 30 recent featured images: 10 product roundups, 15 news posts and five essays. He compared each image with its own headline, a randomly selected other headline and an unrelated plant-care headline. For the product images, he reported that Jev 1.14 gave each image a higher match score against its own headline than against the comparison headlines. Scores for the correct headline ranged from 0.65 to 0.89, with a median of 0.76; random headlines scored no higher than 0.18, and the plant headline no higher than 0.10.
Jev 1.14 · images + box
Jev can see now. Here is what 30 of my images showed.
I asked about every featured image three times: does it fit its own headline, a random other post’s headline, an off-topic headline? Both Jev versions, about 60 requests.
Does this image fit the headline?
Median “yes” probability on Jev 1.14, by kind of post
- Product roundups10 images0.760.061.13: 0.79 / 0.05
- News posts15 images0.330.161.13: 0.51 / 0.25
- My essays5 images0.510.161.13: 0.62 / 0.33
Off-topic headline (“How to Repot a Monstera Plant”): never above 0.10 on either version, 60 of 60 rejected.
Why news is weak: almost all news images show the same scene, people at desks looking at monitors. Jev could not tell them apart, and it was right not to.
1.14 is stricter than 1.13
Same 30 images, same questions
| Result | 1.14 | 1.13 |
|---|---|---|
| Fits own headline (≥ 0.5) | 15/30 | 23/30 |
| Fits a wrong headline (≥ 0.5) | 1/30 | 4/30 |
| Own beats random other (AUC) | 0.91 | 0.89 |
| Names the product (10 options) | 9/10 | 10/10 |
1.14’s one product miss: mesh system read as “Wi-Fi 7 router”, confidence 0.60. Re-tune thresholds before jev-latest moves.
What shipped on Oct 14
- headerImagesin state, instructions or criteria · 1.13 + 1.14 ·
TypeSafe-Danger-Zone-Images - 1.14 onlyBox typeone bounding box + confidence. 6 of 6 clear targets boxed (0.66–0.90); an absent bicycle still got a box (0.03)
- headerLong context64k / 32k → 262k tokens, slower and less accurate
- opt-inHelp find weaknessesqueries used for testing, not training
now: jev-preview = 1.14~Oct 21: jev-latest → 1.14notice: 60 days (1.13), 30 days (1.14)
Source: own test on October 15, 2026. 30 featured images from ThorstenMeyerAI.com (10 product roundups, 15 news posts, 5 essays), JPEG at up to 1,568 px, jev-1.14.0 and jev-1.13.0 with the image header. Bars show medians. AUC = probability that an image’s own headline scores higher than a random other headline. Features, prices and limits: TypeSafe blog post “Jev 1.14” and docs.typesafe.ai, as of October 15, 2026.
Powered by Thorsten Meyer AI
Testing Jev on Real Site Images
The release extends Jev from text-only structured decisions to image-based classification and localization. That could let developers ask bounded questions about visual material—for example, whether an image fits a headline or where a named object appears—without using a general-purpose chatbot to generate an open-ended description.
Meyer’s test offers an early, small-scale example of that use. On the 10 product images, Jev 1.14 selected the correct product in nine cases out of 10. The one error involved a mesh Wi-Fi image showing both a router with antennas and a cylindrical mesh unit; Jev chose “Wi-Fi 7 router” with confidence of 0.60. Meyer said all correct product selections had confidence of at least 0.77, suggesting a threshold could have flagged the miss in this particular sample. That observation is not evidence that the same threshold will work reliably on other image sets.
The feature may be useful for repetitive, narrowly defined checks, but the limits and TypeSafe’s own warning matter for anyone considering deployment. The test involved one site’s images and a limited set of questions. It does not establish performance across image types, subjects, or production workflows.
As an affiliate, we earn on qualifying purchases.
A Cautious Opt-In Release
Before this release, Jev’s typed questions had to concern text, according to Meyer’s account. TypeSafe said in its release post that image support also works on the current model and described jev-1.13.0 as having been multimodal, even though the capability had not been exposed in the same way. Jev 1.14 also raises the optional context limit from 64,000 to 262,000 tokens per request, and from 32,000 to 262,000 tokens per question, through a separate header. TypeSafe advises against the larger limit because reliability can decline with long inputs.
The release keeps these newer functions out of requests unless developers explicitly opt in. TypeSafe described the approach as a way to preserve reliability for production builders while allowing experimentation. The company said there is still “way too much jaggedness” to recommend the features for production use cases. Images carry no separate surcharge, according to the source report; they are billed as regular input tokens at Jev’s stated rate of $0.042 per million input tokens. Base64 encoding adds about one-third to the file size, while the Python and JavaScript SDKs support multipart uploads.
TypeSafe also released Jev 1.14 as jev-preview. Meyer reported that the jev-latest alias was expected to move to the new version about a week after release, around October 21. TypeSafe shared internal intelligence estimates rather than public benchmark scores, and the source says the company cautioned that those comparisons were limited to open-weight models and might favor Jev.
““There is still way too much jaggedness for us to vouch for these for production use cases.””
— TypeSafe, as quoted in Meyer’s account of the release
As an affiliate, we earn on qualifying purchases.
What the Image Test Cannot Show
The supplied source material ends as Meyer begins describing the 15 news-image results, so their outcomes are not available here. It also does not provide the results for the five essay images or a complete comparison of Jev 1.14 with Jev 1.13 across all questions. Meyer says he ran both versions on about 60 requests, but the available account gives detailed results only for the product-image portion.
The test is an individual evaluation of images from one website, not an independent benchmark. It does not establish how Jev performs on other publishers’ images, varied photography, accessibility tasks, or real-world production traffic. TypeSafe itself says the new capabilities remain too unreliable for the company to vouch for in production. The source also does not specify how the model’s confidence scores should be calibrated or whether the observed threshold would generalize.
As an affiliate, we earn on qualifying purchases.
The Jev Rollout and Further Tests
Jev 1.14 is available through the jev-preview model name. According to Meyer’s report, TypeSafe planned to move jev-latest around October 21, roughly one week after the release. The company’s stated deprecation windows were 60 days for version 1.13.0 and 30 days for 1.14.0, with a formal policy still to come.
For users considering the image or bounding-box features, the next practical step is to test them on their own data and narrow tasks rather than treat Meyer’s results as a general performance guarantee. TypeSafe has not said in the supplied material when it expects the experimental features to be ready for production or what reliability benchmarks would change that assessment.
As an affiliate, we earn on qualifying purchases.
Key Questions
What changed in Jev 1.14?
TypeSafe added opt-in image inputs and a bounding-box response type. Image support is also available on Jev 1.13.0, while the box type is limited to version 1.14.
How did Jev perform on the product images?
In Meyer’s test of 10 product images, Jev 1.14 matched each image to its own headline more strongly than to the comparison headlines. It identified the pictured product correctly in nine of 10 cases.
Did the test establish how Jev performed on news images?
No. The source material cuts off as Meyer starts discussing the 15 news posts, so their results are not available in the account provided.
Is TypeSafe recommending image support for production?
No. TypeSafe said the new capabilities still have too much “jaggedness” for the company to vouch for them in production. They require an explicit opt-in header.
What image formats and limits does Jev support?
The reported limits are PNG and JPEG still images, 256 to 1,568 pixels per side, 16 MiB per file and 32 MiB per request. The source says WebP and HEIC files need conversion first.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
