FHE Benchmarking Results - ML Inference (Medium Batch - 1000)

Submitter Bandwidth Quality Timing (harness) Timing (server)
Name Date Env R/L Keys Input Result Accuracy Total DB Gen Keygen Model Prep Input Gen Input Prep Input Enc Compute Decrypt Total DB Gen Keygen Model Prep Input Gen Input Prep Input Enc Compute Decrypt
Reference2026‑02‑20 03:38:16CPUL1G4.9G1001.6M0.9741.8286h4.2678s966ms3.9ms2.0617s186ms35.93s1.8091h24.74s1.7744h1.7744h
Lattica-ai2026‑02‑23 16:49:29GPUR16.5M196M2.5M0.97232.26s7.3829s1.8578s200us3.1782s1.8901s5.1136s3.5153s2.2483s605.3ms