San Francisco – July 28, 2025 — Paravision, a leader in trusted Identity AI, today announced exceptional results in the latest NIST Face Recognition Technology Evaluations (FRTE), confirming its upcoming Gen 7 Face Recognition as the top-ranked technology from the Americas and Europe, and one of the most accurate and consistent solutions globally.

In both the 1:1 (verification) and 1:N (identification) evaluations, Paravision’s Gen 7 model placed among the top 5 global performers in the key “Visa-Border” benchmark. Among vendors from the Americas and Europe, Paravision achieved the #1 accuracy across both tracks.

Consistent Accuracy Across a Range of Real-World Scenarios

In NIST FRTE 1:1, Paravision (as submission paravision-018) earned top 6 rankings in all 7 testing scenarios, including those involving difficult poses, long time gaps, and border use cases. Its 1:N submission (paravision-019, using the exact same models as paravision-018) secured top 10 results in 6 of 8 1:N scenarios, particularly excelling in mugshot, kiosk, and border matching, proving readiness for high-volume deployments. 

These results reflect our dedication to pushing the limits of real-world identity performance. To rank alongside the best globally while leading in transparency, accessibility, and trust is a direct outcome of our team’s focus on both innovation and mission impact.
Doug Aley, Paravision CEO

Consistent Leadership Across Multiple Generations

Paravision Gen 7 leadership in NIST FRTE signals not only accuracy but consistent accuracy over product generations and years of testing. Since Paravision was first ranked #1 globally in NIST FRTE (then FRVT) in 2019, its error rates have decreased 80% for 1:1 Verification and 90% for 1:N Identification. Meanwhile, for the last four product generations, Paravision Face Recognition has ranked in the top 6 globally for both 1:1 and 1:N. As always, Paravision is committed to NIST FRTE testing of its production software and transparency on which submissions correlate to which product versions. 

Six years ago, Paravision was the new kid on the block and surprised everyone when Paravision Face Recognition came out on top in NIST testing. Today, we’re thrilled to say this latest result continues a trend of many submissions over many years, and gives our partners another reason to put their trust in us.
Joey Pritikin, Paravision Chief Product Officer

Continued Performance Gains from Gen 6 to Gen 7

Paravision’s Gen 7 models deliver meaningful gains over their predecessors, especially in critical use cases. For example:

Scenario Gen 6 FNMR → Gen 7 FNMR % Error Reduction
1:1 VISA-BORDER 0.0019 → 0.0016 15.8% 
1:1 BORDER-BORDER 0.0041 → 0.0034 17.1% 
1:1 BORDER-KIOSK 0.0403 → 0.0375 6.9% 
1:N VISA-BORDER 0.0026 → 0.0022 15.4%
1:N MUGSHOT-PROFILE 90° 0.4311→ 0.1439 66.6%
1:N BORDER ΔT ≥ 10 YRS 0.0583 → 0.0450 22.8%

This progress is especially impactful for regulated and border use cases, where even small improvements translate into significantly lower false reject rates at scale.

We built Gen 7 to raise the bar across every dimension: accuracy, pose robustness, and acquisition consistency. These results validate that our core AI research is solving for the complexity of modern identity systems.
Charlie Rice, CTO of Paravision