🔍 Read the full analysis: I Let Jev Examine 30 Images From My Collection on ThorstenMeyerAI.com
Prime made for students and young adults
- Fast, free delivery for dorm and study essentials
- Prime Video and Amazon Music included
- Member-only deals
TL;DR
TypeSafe released Jev 1.14 on October 14, adding opt-in image inputs and a bounding-box response type while warning that the features are not ready for production use. In a test of 30 images from ThorstenMeyerAI.com, Jev 1.14 matched product photos to headlines well and identified 9 of 10 products; the source account is cut off before reporting the news-image results.
Jev takes a state and typed questions, returning answers such as yes-or-no probabilities, selections from a list or scores on a scale. Before this release, its questions were limited to text, Meyer wrote. TypeSafe says image support also works on jev-1.13.0, which the company described as having been multimodal already. Image requests require the explicit header TypeSafe-Danger-Zone-Images: true; a separate opt-in header enables longer context.
The new box response type is available only on Jev 1.14 and returns one bounding box for a specified target, alongside a confidence score. TypeSafe says the image and bounding-box features are still too unreliable for it to endorse for production use, so they are disabled unless a request opts in. The release also raises the optional context limit from 64,000 to 262,000 tokens per request, while retaining the existing limit by default and advising against longer inputs because reliability can fall.
Meyer tested the two model versions on the site’s 30 most recent featured images: 10 product roundups, 15 news posts and five essays. Each image was checked against its own headline, a randomly chosen other headline and an unrelated headline about repotting a monstera. The 10 product images also received a multiple-choice question about which product was shown. Meyer said the test involved about 60 requests across both models.
For product images, Jev 1.14 gave the image’s own headline a median fit score of 0.76, compared with a median of 0.06 for randomly selected other headlines; the unrelated plant headline scored no higher than 0.10. The model picked the correct product in nine of 10 cases. Its one error was a mesh Wi-Fi image, which it identified as a Wi-Fi 7 router with 0.60 confidence. Jev 1.13 identified all 10 products, but 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 Cautious Route to Image Checks
The release extends Jev from text-based classification into image-based decisions, including checks that could be applied repeatedly to editorial assets or product catalogs. Meyer’s test suggests the model can distinguish product photos associated with different headlines in this small sample. That is a useful early signal, not evidence that it will perform reliably on other collections or workflows.
TypeSafe’s opt-in design matters because the company is warning users not to treat the new functions as production-ready. The confidence scores also did not eliminate error: the product-identification miss scored 0.60, while the earlier model had correct answers with confidence below 0.5. Users considering the feature would need their own evaluations and a policy for handling uncertain or incorrect outputs.
The test’s incomplete reporting limits what readers can conclude about news images. The available account says the 15 news posts were part of the evaluation, but it cuts off as Meyer begins discussing their results. No findings for those images can be stated from the supplied material.
product photo recognition AI tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What Jev 1.14 Adds
TypeSafe describes Jev as a decision model that responds to typed questions rather than generating ordinary prose. Its new image format uses an assets map: a request assigns an image to a reference, and that reference can appear in the state, question instructions or criteria. Meyer advises using a fresh UUID for references because matching is plain text and a short string could unintentionally replace part of another word.
Image placement can affect cost: an image in the state is reused across questions, while an image placed in each question’s instructions is sent again for every question that uses it. The release accepts still PNG and JPEG files from 256 to 1,568 pixels per side, with limits of 16 MiB per file and 32 MiB per request. WebP and HEIC files must be converted, and one invalid image can fail the whole request. TypeSafe says image input is billed at the model’s standard rate of $0.042 per million input tokens, with no separate image surcharge.
Jev 1.14 is available as jev-preview. TypeSafe said jev-latest would move to the version about a week after release, around October 21. The company also announced 60 days’ deprecation notice for version 1.13.0 and 30 days for 1.14.0, with a formal policy still to come. Meyer noted that TypeSafe’s internal comparisons use open-weight models and may be biased toward Jev, and said users should run their own evaluations.
“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 quoted by Thorsten Meyer
image analysis bounding box software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
News Image Findings Are Missing
The source account supplied for this report ends just as Meyer begins discussing the 15 news-image results. It does not say how Jev scored those images or whether either version made errors, so no conclusion about performance on news photography is supported here. The test also covered only 30 images from one site and does not establish how the model performs on other subjects or image collections.
TypeSafe has not vouched for image analysis or expanded context in production. The source does not provide independent benchmark results, and the company’s internal comparisons are described by Meyer as potentially biased toward Jev. It remains unclear when TypeSafe will consider the new features reliable enough for production guidance or publish a formal model deprecation policy.
AI image recognition for product photos
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Preview Rollout and Further Tests
TypeSafe said Jev 1.14 was available through jev-preview and planned to move jev-latest to the new version around October 21, 2026, one week after release. Its stated deprecation notice periods are 60 days for Jev 1.13.0 and 30 days for 1.14.0; the company said a formal policy is still forthcoming.
For now, the release leaves production validation to users willing to opt in. Meyer said he ran his own evaluation rather than relying on TypeSafe’s internal model comparisons. The next useful evidence would include the missing news-image results, tests on larger and more varied datasets, and further guidance from TypeSafe on reliability and version support.
As an affiliate, we earn on qualifying purchases.
Key Questions
What changed in Jev 1.14?
TypeSafe added opt-in image inputs and a new bounding-box response type. The box type is limited to Jev 1.14; TypeSafe says image support also works on Jev 1.13.0.
What did the 30-image test find?
For 10 product images, Jev 1.14 matched images to their own headlines with a median score of 0.76 and identified the featured product correctly in nine cases. The supplied source cuts off before reporting the results for the 15 news images and five essays.
Is TypeSafe recommending Jev image analysis for production?
No. TypeSafe said the feature remains too unreliable for it to vouch for production use. Image analysis must be explicitly enabled with a request header.
When does Jev 1.14 become the default latest model?
TypeSafe said jev-latest would move to Jev 1.14 around October 21, 2026, about a week after release. The model was available sooner as jev-preview.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
