OpenJev vs Jev: the same question, two different models
We hosted one of these and never the other, so read this knowing that. The short version: they take the same request and return the same shape, they are unrelated models by unrelated people, we expected ours to be the weaker one, and nobody measured the gap — including us.
OpenJev is retired — here is what replaced it
On 2026-10-07 we stopped serving OpenJev. Our judgment endpoint now speaks OpenAI's Decisions API and answers from gpt-6-luna: the same three question types under new names — predicate, choice, score — over text or images, free with a registered key at https://api.blockrun.ai/v1/decisions.
The old /v1/decide request still works there and is translated. Neither OpenJev nor its replacement is Jev, and we still do not resell Jev. The comparison below is kept as the record of what OpenJev was.
The Decisions endpoint has the request, how to get the free key, and the failure we measured on the new model.
The names are the problem
OpenJev is not an open version of Jev. It is not a smaller checkpoint, a free tier, or a community port. It is a separate MIT-licensed model that took the name because it answers the same kind of question.
If you arrived expecting the open weights behind TypeSafe's model, they are not here and they are not anywhere. We do not resell Jev; TypeSafe sells it directly.
Side by side, only where a comparison is real
| OpenJev | Jev | |
|---|---|---|
| Who makes it | AlexWortega, open source | TypeSafe |
| Licence | MIT | Commercial, theirs |
| Where we serve it | Nowhere since 2026-10-07 — retired | We never served it |
| Request shape | state + typed questions | state + typed questions — the same shape |
| Question types | yes/no, choice, score | yes/no, choice, score |
| What the number is | A share of the entailment found across your options | Probabilities fitted against real outcomes |
| Input | Text and JSON; the checkpoint also read images, we never exposed that | Text |
| Quality against each other | Not measured by anyone, including us | Not measured by anyone, including us |
The last row is the one most comparison pages would quietly fill in. No published head-to-head exists, we have not run one, and a figure invented to fill that cell would be the least trustworthy thing on the page.
The one difference that is concrete
Forget which is "better" for a moment. There is a difference in what the numbers mean, and it changes how you are allowed to use them.
Jev is trained so its probabilities line up with outcomes: across many predictions, the ones it calls seven-in-ten happen about seven times in ten. That is what calibration buys you, and it is why you can set a threshold from the number itself.
OpenJev was an entailment model, and for a labelled choice its scores were divided by their total so they summed to one. Three options it supported equally weakly and two it supported equally strongly produced the same even split. The number was a share of the agreement found, not a probability the answer was right, so a threshold had to come from your own labelled examples instead of from the number's face value.
That was not a quality claim. It was a difference in kind.
Which to use
When the probability has to mean something — a threshold you will defend, a risk budget, anything where "act above 0.85" is a policy rather than a hunch. Also when you want the strongest available model for this task and are willing to pay for it. Go to TypeSafe.
OpenJev is no longer an option. When you want to find out whether typed judgments help at all, or free and no-signup-beyond-a-key matters most, our endpoint now answers from gpt-6-luna on the same key. We have not compared it with Jev either, and will not put a number on that.
A third option people skip: try both against a chat model with a JSON schema on your own data. On BlockRun that is one key and a changed string, which is the whole reason this comparison is worth our time to write.
Read next
- Decisions — the endpoint that replaced OpenJev, the three question types, and how to call it free.
- API reference — the request shape and the response fields.
OpenJev vs Jev questions: differences, retirement, what to use now
- OpenJev vs Jev — what was the difference?
- They were unrelated models by unrelated people that took the same request shape. Jev is TypeSafe's commercial model; OpenJev was MIT-licensed and open. The concrete difference was what the numbers meant, not just who made them.
- Is OpenJev still available on BlockRun?
- No. We retired it on 2026-10-07. Our judgment endpoint now speaks OpenAI's Decisions API and answers from gpt-6-luna, free with a registered key at api.blockrun.ai/v1/decisions. The old /v1/decide request still works there and is translated to the new one.
- Was OpenJev as good as Jev?
- We expected it to be materially weaker, and we never measured the gap. No published head-to-head exists and we did not invent a figure to fill that cell — we hosted one of the two, which is exactly why we would not.
- What does calibration mean when comparing Jev and OpenJev?
- A calibrated model's probabilities line up with outcomes: across many predictions the ones it calls seven-in-ten happen about seven times in ten, so a threshold can come from the number itself. OpenJev's choice scores were entailment shares normalised to sum to one, so a threshold had to come from your own labelled examples instead.
- When should I use Jev?
- When the probability has to mean something — a threshold you will defend, a risk budget, a policy rather than a hunch — or when you want the strongest available model for the task and will pay for it. TypeSafe sells it directly.
- What replaced OpenJev, and is it Jev?
- gpt-6-luna through OpenAI's Decisions API, with the same three question types under new names: predicate, choice and score. It is not Jev, and we have not compared it with Jev either.
- Can I get Jev through BlockRun?
- No. We do not resell it and there is no gated path waiting to be switched on.
- Were OpenJev's weights the open version of Jev?
- No, and this was the most common wrong assumption. TypeSafe's weights are not public anywhere. OpenJev was a separate model that adopted a similar name.
- Why does a page comparing them come from a company that hosted one?
- It should make you read it more carefully. The page conceded the things that cost us something — that we expected ours to be weaker, that nobody measured it, and that Jev is the right answer when the probability has to be defensible — because a comparison from an interested party is worth nothing unless it can be caught being fair.