Disclaimer

VetEval publishes benchmark measurements of AI models. Nothing on this site is veterinary advice, and no score certifies, approves, or licenses any model for clinical use.

1. Research and benchmarking purpose only

VetEval measures how AI models perform on veterinary examination items under fixed, published conditions. A score describes how a model performed on our item set, under our harness and scoring, on a given date. It is not a certification, accreditation, endorsement, warranty, or guarantee of any model's accuracy, safety, or fitness for any purpose, clinical or otherwise.

2. Model outputs may be wrong or unsafe

Any model output shown on this site, including transcripts and examples, may be incorrect, incomplete, misleading, or unsafe. Outputs are reproduced to show what a model did, not because they are correct. Nothing displayed here is veterinary advice or a substitute for the judgement of a licensed veterinarian. Do not rely on anything on this site to diagnose, treat, medicate, or make any decision about the care of an animal. In an emergency, contact a qualified veterinary professional.

3. Not a medical device

VetEval is not a medical device, is not clinical decision-support software, and has not been cleared, approved, certified, or registered by any regulator. It publishes measurements, not clinical guidance.

4. What a score does and does not tell you

• Point in time and version specific. A score reflects the specific model version, endpoint, and settings tested in a specific run. Models change; a score does not update itself.
• Comparable only within the benchmark. Scores are comparable with other VetEval results measured under identical conditions. They are not comparable with numbers produced elsewhere, or across methodology versions, because items are re-weighted when the configuration changes.
• A measurement, with uncertainty. Published intervals and tiers exist because differences smaller than the interval are not real differences. Do not read a rank gap into a difference the interval does not support.
• Examination items, not practice. Performance on examination-style items is not the same as safe, useful behaviour in a live clinical workflow.
• Jurisdictional scope. VetEval reflects a US and NAVLE-aligned standard of care. Jurisdictions with different formularies, licensing rules, or referral norms may weight items differently, and a score may be less representative there.
• Safety findings are bounded. A safety-gate result describes harms observed on our items in that run. The absence of an observed harm is not evidence that a model is safe.

5. Buyers and procurement

If you are selecting an AI model for a veterinary setting, treat VetEval as one input among several. It does not replace your own validation, clinical governance, regulatory assessment, or professional judgement, and viggoVet is not responsible for procurement or clinical decisions made in reliance on a score.

6. Third-party names

Model, product, and company names are used nominatively to identify what was evaluated. Their use does not imply any affiliation with, sponsorship by, or endorsement from those parties, in either direction.

7. Liability

To the fullest extent permitted by law, viggoVet is not liable for decisions made in reliance on any model output, score, ranking, or analysis published on this site. This Disclaimer is subject to the limitations and exclusions in our Terms of Service. Nothing here excludes liability that cannot be excluded by law.