🔍 Read the full analysis: What Jev Made Of 30 Images From My Collection on ThorstenMeyerAI.com
Get everyday essentials 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’s October 14 Jev 1.14 release added opt-in image input and a bounding-box response type, while warning that the new capabilities are not ready for production use. In a test reported by Thorsten Meyer, Jev matched product images to their headlines and identified 9 of 10 products on the newer model; the supplied article cuts off during its account of results for news images.
Thorsten Meyer tested TypeSafe’s Jev decision model on 30 images from his website after the company added opt-in image support and a bounding-box response type in Jev 1.14, released October 14. Meyer reported strong results on product images, but TypeSafe cautioned that the new capabilities remain too unreliable for production use without further evaluation.
Meyer used the site’s 30 most recent featured images: 10 from product roundups, 15 from news posts and five from his own essays. He asked whether each image fit its own headline, a randomly selected other headline and an unrelated plant-care headline. The product images also received a multiple-choice question asking which of 10 products each showed. Meyer says he ran the tests on both Jev 1.14 and Jev 1.13, for about 60 requests overall.
For product roundups on Jev 1.14, Meyer reported that each image scored between 0.65 and 0.89 for its own headline, with a median of 0.76. Random other headlines scored no higher than 0.18, with a median of 0.06; the unrelated plant headline never scored above 0.10. These are the results of Meyer’s limited test, not a public benchmark or a general measure of model performance.
On the product-identification question, Jev 1.14 named the correct product in nine of 10 images, according to Meyer. The error involved a mesh Wi-Fi image showing an antenna-equipped router beside a cylindrical mesh unit: Jev called it a Wi-Fi 7 router, with 0.60 confidence. Meyer says all correct answers had confidence of at least 0.77. Jev 1.13 identified all 10 products, 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
A New Test for Image-Based Decisions
The release extends Jev beyond text-only questions and makes it possible to ask a model for structured judgments about images, rather than asking it to compose a free-form description. That could suit narrow tasks such as checking whether a featured image matches a headline or identifying a pictured product. Meyer’s experiment offers an early example, but its 30 images come from one website and do not establish how the feature will perform across other image types or real production workloads.
TypeSafe’s opt-in design also matters for anyone considering the feature in an application. The company says the model still has too much “jaggedness” to endorse the new functions for production. Keeping them behind explicit request headers means existing behavior remains unchanged unless a developer chooses to test them. That makes the release an experiment for builders, not a claim that image decisions are dependable by default.
As an affiliate, we earn on qualifying purchases.
What Jev 1.14 Adds
Jev is a decision model that answers typed questions about supplied information: it can return a yes-or-no probability, choose from options or score against a scale. Meyer described it in September as a model that does not write sentences. Before the new image feature, he says, its questions had to concern text.
TypeSafe says images can now be included in the state, question instructions or the criteria used to define options and score levels. Image support also works on jev-1.13.0, according to the release post. The new box type, which returns one bounding box and a confidence for a described target, is limited to Jev 1.14. Images must be PNG or JPEG stills, with each side between 256 and 1,568 pixels. TypeSafe lists limits of 16 MiB per file and 32 MiB per request.
The company also added an opt-in long-context header, raising the stated limits from 64,000 tokens per request and 32,000 per question to 262,000 for both. TypeSafe continues to advise against the larger limits because reliability falls on long inputs. Jev 1.14 is available as jev-preview; TypeSafe said jev-latest would move to it about a week after release, around October 21.
““There is still way too much jaggedness for us to vouch for these for production use cases.””
— TypeSafe, in its Jev 1.14 release post
As an affiliate, we earn on qualifying purchases.
The News-Image Results Are Missing
The source provided for this report cuts off immediately after the heading “News images: Jev was right, my images weren’t”, before explaining the test results for the 15 news images. It is not possible to report what Jev scored on those images, how the two model versions compared on them, or what Meyer meant by saying the images were at fault. The article’s product-image findings should not be treated as a substitute for those missing results.
TypeSafe has not endorsed the new features for production, and the account provided does not include independent testing or a broader evaluation. The company has also said its internal model comparisons are estimates rather than public benchmark scores and may favor Jev. The release does not establish how image performance varies across other tasks, image formats or larger sets of examples.
As an affiliate, we earn on qualifying purchases.
Preview Rollout and Further Testing
TypeSafe said it would move jev-latest to Jev 1.14 around October 21, one week after release, while making the new image and long-context behavior dependent on explicit headers. Developers considering the features can test them through jev-preview and run their own evaluations, following TypeSafe’s warning about reliability.
Meyer’s report would need to be completed or supplemented with the missing news-image results before readers can assess his full 30-image experiment. TypeSafe has stated deprecation notice periods of 60 days for version 1.13.0 and 30 days for 1.14.0, with a formal policy still to come. It is not clear from the source whether the company plans additional reliability guidance or a public benchmark.
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 bounding-box response type. The release also includes broader model improvements and an opt-in setting for longer context.
What did Meyer’s image test find?
For 10 product images, Meyer reported that Jev 1.14 matched each image more strongly to its own headline than to unrelated headlines and identified nine of the 10 products correctly. The supplied source does not include the results for the 15 news images.
Is TypeSafe recommending Jev image support for production?
No. TypeSafe said the new capabilities remain too unreliable for the company to vouch for in production use cases. Image support requires an explicit opt-in header.
Does image support work on Jev 1.13?
TypeSafe says image support also works on jev-1.13.0. The new bounding-box type is available only with Jev 1.14.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
