Runs the real public Shenava FastConformer ONNX model on this device, measures decode latency + real-time factor (RTF) per available runtime tier. Works offline-capable everywhere β including no-WebGPU / no-SIMD "toaster" devices.
Tap βRun benchmarkβ. First run downloads the model (then browser-cached).
| Runtime tier | load ms | median decode ms | p90 ms | RTF (Γrealtime) | note |
|---|