ResearchBenchmarksTransformer Inference Latency Estimation on GPT-2 1.5BFollow60.048Latency (s)Bifrost40.22816174.01208307.796441.57992Jun 16, 2026Evaluation ResultsMethodMethodLinksLatency (s)SpeedupBifrostSystem Type=Bifrost, E...System Type=Bifrost, Evaluation Protocol=Euston-style estimator projection2026.0660.0489.252EustonSystem Type=Pure FHE,...System Type=Pure FHE, Evaluation Protocol=Euston-style estimator projection2026.06555.544—