This benchmark contains several interesting workloads for running AI models and machine learning tasks on the client side. The workloads use browser technologies such as WebAssembly and WebGPU and are built on top of popular machine learning frameworks and models. Please see the README for more details.

The frameworks and models used in this benchmark contain assets of considerable size. Currently, the benchmark downloads more than 2 GB of data. Do not run this on a metered connection; a stable broadband connection is highly recommended. Additionally, the first iteration of the benchmark may take several minutes to complete.

Version: 0.2.0 (2026-04-08) (commit: dd536ca)

Your browser window is too small. For most accurate results, please make the view port size at least 850px by 650px.
It's currently .

Score


One or more subtests produced no duration.
Please check your browser settings and re-run the benchmark.

Detailed Results 0.2.0 (2026-04-08)

Non-standard Parameters

Benchmark ran with non-standard parameters.
The results are likely not comparable to default runs.

Param Value Default

Aggregate Metric

Test Metrics Overview


Test Metrics Details