How does ModelVerify.ai check model identity? | The platform runs diagnostic probes against the endpoint and compares the responses to official model fingerprints collected from known-genuine endpoints. Verdicts should be interpreted alongside latency, stability, and drift signals.
Is my API key safe? | Yes. Your API key is stored transiently in Redis, used only for the current run, and deleted immediately after the run completes. It is never written to PostgreSQL or the file store.
What models can I check? | Any model accessible via an OpenAI- or Anthropic-style API endpoint for which the platform has official fingerprints. The official fingerprint model list is available on the verification form.
How accurate is the verification? | Verification accuracy depends on fingerprint coverage, endpoint stability, and model behavior. Results should be interpreted alongside risk levels, drift scores, and stability metrics rather than as a single yes-or-no decision.
What do benchmark scores mean? | Benchmark scores reflect model capability on the baseline_150 test set (0.00-1.00). Higher scores indicate better performance across reasoning, math, code, and QA. Scores are grouped into strong (≥0.85), moderate (≥0.70), and weak (<0.70) bands.
Can I trust third-party API providers? | Model verification provides evidence about whether an endpoint behaves as claimed, but is not a guarantee. Combine verification results with provider reputation, latency, and your own testing.
How to interpret mismatch verdicts? | A mismatch verdict indicates the endpoint behavior differs significantly from the official fingerprint. Check the risk level, drift score, and stability metrics. A single mismatch does not necessarily mean the provider is fraudulent — it may indicate model version differences or configuration issues.