Plans
A free account evaluates one model, once, and that result publishes in full and permanently, which is what makes the benchmark worth citing. Paid plans exist to let you see your score before the world does, evaluate as many models as you like, and run them again.
No vendor funds VetEval’s operations or influences any score. Evaluation fees are uniform and published; they cover compute and grading. Publication is free and score-blind.
Community
Free foreverFor open science. Submit publicly, get ranked, cite the result.
- 1 model, 1 evaluation, and 1 person
- Results always published, in full
- Each result is checked by our team before it goes on the board
- Practice dataset access
No payment details needed
Lab
AnnualFor model teams that need to see scores before the world does.
- 3 private evaluations a year, across all your models
- 6 public evaluations a year, on top of those
- Buy more private evaluations whenever you need them, 2 free public ones come with each
- You decide whether each result is published
- Public results go on the leaderboard as soon as the evaluation finishes
- Unlimited models, and 5 people
Startups get 50% off, apply here
Compare
| Community | Lab | |
|---|---|---|
Models | 1 | Unlimited |
Evaluations you supply the endpoint and key, so the inference is billed to you, not to us | 1 evaluation, once | 6 public evaluations a year, plus every private one you buy |
Private evaluations buy more at any time, and every one you buy carries 2 free public evaluations with it | None | 3 per year, across all your models |
Public evaluations these publish to the board, and they never touch your private allowance | 1, once | 6 per year, across all your models |
People an invitation holds a seat until it is accepted or revoked | 1 | 5 |
Review before publishing the free evaluation is the only one we review | Held for our review before it goes on the board | Goes on the board as soon as the evaluation finishes |
Results | Always published, in full | You decide whether each result is published |
Add-ons
Add-ons increase capacity: more evaluations across the models you have already submitted. They never raise the number of attempts allowed on a single submission, which is capped by methodology, not by price.
Three more private evaluations, usable across any of your models, and 6 free public evaluations with them. A pack does not expire with your billing period, and neither do the public ones.
One more private evaluation, for when a pack is more than you need, with 2 free public evaluations attached.
What money cannot buy
- Rank. Scores come from the evaluation, and nothing else.
- Removal of a published result. Community results are permanent.
- A second published score for a name already on the board. Improve the model, give it a new name, and evaluate it again.
- Sight of the questions. No plan shows an item before a run or after one.
Pricing is uniform and published to every customer. We rank the organizations that pay us, so we do not negotiate the price of a standard plan.
Need something different?
Invoice or PO billing, a multi-year term, more than five models, a security review, or you are a clinic group evaluating which model to buy.