NIST is the U.S. National Institute of Standards and Technology, widely seen as the world’s most recognized authority in biometric benchmarking. Its FRTE 1:1 face verification test is the gold standard for accuracy.
InterLink Labs’ latest model has been officially ranked in the top 51 on NIST’s leaderboard, and the error rates behind that rank matter more than the number itself.
Achieving this requires three things that only a small number of companies with substantial resources can possess: a highly capable team of engineers and researchers to develop the algorithms, a massive amount of training data accumulated over many years, and an annual investment of millions of U.S. dollars in NVIDIA H100 GPUs at that time. Interlink submitted evidence of all these capabilities to NIST and undergo a very rigorous evaluation process.
InterLink's algorithm is ranked 51st in these results. Very few large technology groups are ranked alongside it, including Samsung, Toshiba, SK Telecom, Kakao and Canon.
Fewer bots: a low false match rate keeps fake accounts out of airdrops, rewards and votes.
Verify once, stay verified: stable performance over time means no repeated onboarding.
Fair access worldwide: small gaps across groups means anyone can verify with equal ease.
Privacy by design: face features are extracted on your phone and turned into encrypted, non-reversible templates. Raw face images never leave the device, and a compromised template can be revoked like a password.
The next model is in develpment, aiming for lower latency, better aging performance and tighter fairness. InterLink also plans a public dashboard with live error rates and a model card, plus quarterly metrics reports.
Trust should be measured, not marketed. InterLink’s top 51 ranking is independent proof that the system blocks fake accounts and reliably recognizes real people.
NIST is the U.S. National Institute of Standards and Technology, widely seen as the world’s most recognized authority in biometric benchmarking. Its FRTE 1:1 face verification test is the gold standard for accuracy.
InterLink Labs’ latest model has been officially ranked in the top 51 on NIST’s leaderboard, and the error rates behind that rank matter more than the number itself.
Achieving this requires three things that only a small number of companies with substantial resources can possess: a highly capable team of engineers and researchers to develop the algorithms, a massive amount of training data accumulated over many years, and an annual investment of millions of U.S. dollars in NVIDIA H100 GPUs at that time. Interlink submitted evidence of all these capabilities to NIST and undergo a very rigorous evaluation process.
InterLink's algorithm is ranked 51st in these results. Very few large technology groups are ranked alongside it, including Samsung, Toshiba, SK Telecom, Kakao and Canon.
Fewer bots: a low false match rate keeps fake accounts out of airdrops, rewards and votes.
Verify once, stay verified: stable performance over time means no repeated onboarding.
Fair access worldwide: small gaps across groups means anyone can verify with equal ease.
Privacy by design: face features are extracted on your phone and turned into encrypted, non-reversible templates. Raw face images never leave the device, and a compromised template can be revoked like a password.
The next model is in develpment, aiming for lower latency, better aging performance and tighter fairness. InterLink also plans a public dashboard with live error rates and a model card, plus quarterly metrics reports.
Trust should be measured, not marketed. InterLink’s top 51 ranking is independent proof that the system blocks fake accounts and reliably recognizes real people.