🔍 Read the full analysis: How 30 Personal Images Helped Me Explore Jev’s Vision 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 input and a bounding-box answer type. In a test of 30 images from ThorstenMeyerAI.com, the author reported strong separation for product images, but the supplied source cuts off before reporting results for news images, essays or the older model comparison in full.
TypeSafe added opt-in image input and a new bounding-box response type to its Jev decision model in a release published October 14, 2026. In a test published the next day, Thorsten Meyer of ThorstenMeyerAI.com said Jev 1.14 matched product images to their own headlines and identified nine of 10 pictured products, while TypeSafe warned that the new capabilities were not ready for production use.
Jev is designed to answer structured questions rather than generate prose: it can return a yes-or-no probability, choose among options or score against a scale. Before the release, those questions had to be based on text. TypeSafe says images can now be placed in a request’s state, question instructions or criteria for options and scores. The feature works with jev-1.13.0 and jev-1.14.0, but requires the header TypeSafe-Danger-Zone-Images: true.
The release also introduced a box type in Jev 1.14, which returns one bounding box for a described target along with a confidence score. A separate opt-in header raises the request context limit from 64,000 tokens to 262,000 and the per-question limit from 32,000 to 262,000. TypeSafe says longer inputs can reduce reliability and advises against using the expanded limits without a reason.
Meyer tested the image feature with 30 recent featured images from his site: 10 product roundups, 15 news posts and five essays. He sent each image with questions about its fit 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’s score for the matching headline ranged from 0.65 to 0.89, with a median of 0.76. Scores for other headlines reached no more than 0.18, and the unrelated headline scored no higher than 0.10.
On a separate question asking which of 10 products appeared in each product image, Meyer said Jev 1.14 identified nine correctly. The one error was a mesh Wi-Fi image that showed both a router and a cylindrical mesh unit; Jev selected “Wi-Fi 7 router” with 0.60 confidence. Meyer said correct answers had confidence scores of at least 0.77. Jev 1.13 identified all 10 products in his test, though two answers had confidence below 0.5.
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
What the Image Test Shows
The results point to a possible use for Jev in repetitive image checks, such as comparing a product photo with a headline or selecting an item from a defined list. Meyer’s test is a small, site-specific evaluation, not an independent benchmark. Its product-image results show how the model behaved on this sample, but do not establish accuracy across other image types, publishers or real-world workflows.
TypeSafe’s opt-in approach also affects how developers can assess the feature. Image requests do not change production behavior unless callers set the danger-zone header, according to the release description. That makes experimentation possible without automatically applying the new behavior to existing requests. The company’s warning that there is too much “jaggedness” for production use remains a clear limit on how confidently the results should be applied.
For readers evaluating AI image tools, the reported confidence scores are not a substitute for checking errors. The product test included a miss at 0.60 confidence, while Jev 1.13’s correct product identifications included two scores below 0.5. The sample therefore does not establish a reliable confidence threshold that would catch every mistake.
As an affiliate, we earn on qualifying purchases.
Jev’s Shift Beyond Text
Meyer previously described Jev as a model for structured decision tasks and wrote about its potential uses in September. Until this release, users supplied text-based state and typed questions, then received constrained answers rather than open-ended prose. TypeSafe says the current model, Jev 1.13, was “multimodal all along,” although image input had not previously been exposed in the described workflow.
TypeSafe framed the release as a choice between reliability for builders and experimentation for users. The company says existing production behavior remains unchanged unless developers send the relevant headers. Image files must be PNG or JPEG stills, with each side between 256 and 1,568 pixels; a file may be up to 16 MiB, and a request up to 32 MiB. The documentation described in Meyer’s report does not provide automatic resizing, and one invalid image can fail an entire request.
Jev 1.14 is available as jev-preview. TypeSafe said jev-latest would move to that version about one week after release, around October 21. The company also shared internal intelligence estimates rather than public benchmark scores, and cautioned that comparisons with open-weight models may favor Jev.
“was multimodal all along”
— TypeSafe, describing Jev 1.13.0 in the release post, as quoted by Meyer
As an affiliate, we earn on qualifying purchases.
Results Missing From the Report
The source material provided for this article ends just as Meyer begins discussing the 15 news images, with the sentence “The news posts were a dif”. It does not include the remainder of his results. Findings for the five essays, the comparison between Jev 1.14 and Jev 1.13 on the headline-fit questions, and the full total of roughly 60 requests are also not available in the supplied text. Those outcomes cannot be inferred from the product-image results.
It is also unclear how well the reported findings would generalize beyond the 30 images on Meyer’s site. The article describes the sample and questions, but the available source does not establish an external evaluation, a broader benchmark or production performance. TypeSafe itself says it cannot yet vouch for these capabilities in production.
As an affiliate, we earn on qualifying purchases.
Release Schedule and Further Testing
TypeSafe’s stated next release step is moving jev-latest to Jev 1.14 about a week after the October 14 launch, around October 21, 2026. The company also set out deprecation notice periods: 60 days for Jev 1.13.0 and 30 days for Jev 1.14.0, while saying a formal policy is still to come.
For developers, the immediate next step is testing the opt-in capabilities against their own images and decision tasks rather than treating Meyer’s product results as a general guarantee. TypeSafe says users can opt in to share queries to help identify model weaknesses, while stating that it will not train on those queries. The timing and details of any further reliability updates are not specified in the supplied material.
As an affiliate, we earn on qualifying purchases.
Key Questions
What changed in Jev 1.14?
TypeSafe added opt-in image input and a new box response type that returns a bounding box and confidence score. The release also added an opt-in longer-context setting.What did the 30-image test find?
For 10 product images, Meyer reported that Jev 1.14 gave each image’s own headline a higher fit score than the tested alternatives. It identified nine of the 10 products correctly. The supplied source does not include the results for the news images or essays.Is Jev’s image feature ready for production?
TypeSafe said it was not yet willing to vouch for the new capabilities in production use cases. Image support requires an explicit danger-zone header.Which image files does Jev accept?
The release details in Meyer’s report specify PNG and JPEG still images, between 256 and 1,568 pixels per side. WebP and HEIC files need conversion first.When does Jev 1.14 become the default latest version?
TypeSafe said jev-latest would move to Jev 1.14 about one week after the October 14 release, around October 21, 2026.Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
