🔍 Read the full analysis: How Well Does Jev See? I Tested It On My Own Images on ThorstenMeyerAI.com
Turn your quiet moments into listening time
- Thousands of audiobooks, podcasts and originals
- Listen on your phone, tablet or Echo — also offline
- Cancel anytime
TL;DR
TypeSafe released Jev 1.14 on October 14, 2026, adding opt-in image input and a bounding-box response type to its decision model. In a test of 30 images published by ThorstenMeyerAI.com, Jev 1.14 matched product images to their own headlines and identified 9 of 10 products, while the supplied account cuts off before reporting results for news and essay images.
Jev is a decision model that responds to structured questions with outputs such as yes-or-no probabilities, choices from a list or scores. Before the release, questions had to be about text. TypeSafe says images can now be included in a request’s state, question instructions or criteria, using an explicit opt-in header. Image support also works with Jev 1.13.0, according to the company.
Jev 1.14 adds a box response type that returns one bounding box and a confidence score for a specified target. Both image use and box responses are marked as experimental. TypeSafe said it sees too much “jaggedness” to vouch for them in production use cases, so they require explicit opt-in headers rather than changing the default behavior.
Meyer tested 30 featured images from his site: 10 product roundups, 15 news stories and five essays. Each image was used in three yes-or-no checks: whether it fit its own headline, a randomly selected different headline and an unrelated headline about repotting a monstera plant. Product images also faced a question asking Jev to identify which of 10 products each showed. Meyer ran the tests on both Jev 1.14.0 and 1.13.0, reporting about 60 requests overall.
For product images, Meyer reported that Jev 1.14 distinguished each image’s own headline from the comparison headlines. Own-headline scores ranged from 0.65 to 0.89, with a median of 0.76. Random other headlines scored no higher than 0.18, with a median of 0.06; the plant headline scored no higher than 0.10. Jev 1.14 named the correct product in 9 of 10 images. The older model named all 10 correctly, 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
Testing Jev Against Real Site Images
The test offers a small, practical example of how image input could support repetitive classification tasks, such as checking whether a featured image suits a headline or identifying a product in a picture. Those tasks are narrower than open-ended image description and fit Jev’s stated role as a structured decision model.
The results also show why TypeSafe has kept the new functions behind opt-in headers. Jev 1.14 made one product-identification error, and the article’s available test account does not report the outcomes for its 15 news images or five essays. A 30-image sample from one website is not enough to establish how the model will perform across other image types, publishers or production settings.
For readers considering the model, the headline and product results are evidence to test the feature, not a broad performance guarantee. Meyer’s comparison also shows that confidence scores may help flag uncertain answers, but the reported results do not establish a dependable threshold for other tasks.
As an affiliate, we earn on qualifying purchases.
What TypeSafe Changed in Jev
Jev’s format centers on a supplied state and typed questions rather than composing prose. In its October 14 release, TypeSafe added image references that can be placed in a request’s state, instructions or answer criteria. The company says images are billed as ordinary input tokens at Jev’s usual rate of $0.042 per million input tokens, with output tokens free.
The release also includes a separate opt-in for longer requests. TypeSafe says the default limits are 64,000 tokens per request and 32,000 per question; the header raises both to 262,000. The company retained the lower default and advised against relying on the larger context because reliability can fall on long inputs.
TypeSafe made Jev 1.14.0 available as jev-preview. According to the release details provided by Meyer, the jev-latest alias is due to move to that version about a week after release, around October 21. TypeSafe also said it plans deprecation notices for older versions, though a formal policy is still to come.
““There is still way too much jaggedness for us to vouch for these for production use cases.””
— TypeSafe, as quoted in the October 15 account by Thorsten Meyer
As an affiliate, we earn on qualifying purchases.
Results Beyond Product Photos
The supplied source account ends mid-sentence as it begins discussing the 15 news images. It does not provide the news-image scores, the results for the five essays, or a full breakdown of Jev 1.13 versus Jev 1.14 across all image categories. Those outcomes cannot be inferred from the product-photo results.
The test also does not establish performance on a representative dataset. It used a small sample from one website, and the account does not provide details such as repeated-run consistency, independent grading or a formal benchmark. TypeSafe’s internal estimates, Meyer says, compare Jev with open-weight models and may be biased toward Jev; they are not public benchmark scores.
Technical constraints also apply: the release supports still PNG and JPEG files within specified size limits, while WebP and HEIC images need conversion. TypeSafe’s warning leaves open how reliably the experimental features will perform in real production workloads.
As an affiliate, we earn on qualifying purchases.
Release Rollout and Further Testing
TypeSafe’s stated schedule is to move jev-latest to Jev 1.14.0 around October 21, 2026, one week after its release as jev-preview. Developers can test the preview sooner, but the company’s opt-in requirement and reliability warning remain in place.
TypeSafe has said it will provide deprecation notice for earlier versions, with a formal policy still pending. For now, Meyer’s report points to running task-specific evaluations before relying on image outputs. A fuller account of the tests—especially the missing news and essay results—would be needed to judge how consistently Jev handles image-to-headline matching beyond product photos.
bounding box object detection software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did TypeSafe add in Jev 1.14?
TypeSafe added opt-in image input and a new box response type that returns one bounding box and a confidence score. The release also offers an opt-in longer-context setting.
How did Jev perform on product images?
In Thorsten Meyer’s test, Jev 1.14 identified the right product in 9 of 10 images and gave higher headline-fit scores to each product image’s own headline than to the comparison headlines. These are results from a 10-image sample on one website.
Did Jev 1.13 perform better than Jev 1.14?
On the product-identification question, Jev 1.13 got all 10 answers right, while Jev 1.14 got nine right. Meyer said two of the older model’s correct answers had confidence below 0.5. The available account does not give complete version-by-version results for all image categories.
Can Jev’s new image features be used in production?
TypeSafe has not vouched for the features in production use cases, citing reliability issues. Image support and bounding boxes require explicit opt-in headers, and developers should test them against their own tasks.
When will Jev 1.14 become the default latest version?
TypeSafe’s stated plan is for jev-latest to move to Jev 1.14.0 around October 21, 2026, about a week after the October 14 release. Jev 1.14.0 is available as jev-preview in the meantime.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
