Short version: we collect anonymous GPU benchmark stats. Nothing personal.
When you run benchmarks, we save:
Collected data builds an aggregate picture of WebGPU compute and transformer inference performance across real-world hardware. We may publish aggregate statistics in research preprints or public reports. Individual submissions are never published or shared.
All benchmark shaders run entirely on your GPU inside your browser. No computation is offloaded to our servers. Only the final results are transmitted after benchmarks complete.
Benchmark results are stored indefinitely as anonymous records. Because no personal information is attached, individual records cannot be identified or deleted.
Last updated: March 2026
Questions? GitHub →