Platform
What it takes to run.
Measured figures, each with the machine and the conditions it was measured under. Where a number was estimated rather than measured, it is not on this page.
On a Raspberry Pi Zero 2 W
The whole engine was run on a physical Pi Zero 2 W, a board with 416 MB of usable memory, on 21 September 2026.
engine binary, compiled for arm64
Pi Zero 2 W, 2026-09-21
peak memory, while holding a 651 MB knowledge graph on disk
Pi Zero 2 W, 2026-09-21
to answer a question, once warm
Pi Zero 2 W, 2026-09-21
sent off the board during the run
Pi Zero 2 W, 2026-09-21
Judging web traffic
On one processor core, judging the requests of the CSIC 2010 web-attack dataset, the engine added about 0.09 ms to each request, handled about 11,000 requests a second and peaked at 29 MB of memory.
How it computes
Senua AI uses no floating-point arithmetic: every value is a fixed-point integer, so a result is the same on every machine. It uses no dense matrix multiplication and needs no GPU. Every public showcase is served by one small cloud machine with 2 processor cores and 2 GB of memory.