FHE Benchmarking Results - MNIST Inference (Single)

Submitter					Bandwidth			Timing (harness)									Timing (server)
Name	Date	Env	R/L	Model	Keys	Input	Result	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
Lattica-ai	2026‑02‑23 16:49:29	GPU	R	mlp	113M	256.1K	256.1K	26.79s	7.3174s	3.404s	200us	2.5856s	1.6016s	1.6353s	1.8656s	1.6942s								48.33ms
Lattica-ai	2026‑05‑25 08:42:44	GPU	R	mlp	46M	128.1K	128.1K	24.45s	7.1235s	2.1834s	200us	2.5703s	1.5982s	1.6326s	1.8617s	1.6477s								15.33ms
CryptoLab Inc.	2026‑09‑07 06:57:38	GPU	L	mlp	108.5K	51.2M	350.1K	7.2068s	3.7867s	147.5ms	79.2ms	1.84s	62.93ms	496.9ms	417.4ms	313.3ms	89.41ms							567us
Reference	2026‑02‑20 03:38:16	CPU	L	mlp	1G	5M	1M	15.57s	4.7884s	975.1ms	3.7ms	1.9934s	3.5ms	92.47ms	7.6306s	83.93ms	6.48s							6.48s