For Codex environment comparisons

Find out what your browser GPU path is actually doing.

Run the same local WebGPU workload in each environment. The report separates real browser support and measured dispatch timing from conclusions about Codex desktop or CLI.

WEBGPU ADAPTER LIMITS   /   COMPUTE DISPATCH SAMPLES   /   JSON EVIDENCE   /   WEBGPU ADAPTER LIMITS   /   COMPUTE DISPATCH SAMPLES   /   JSON EVIDENCE   /  

Run the same test everywhere

Probe this browser, then take the result with you.

Workload controls

Ready to inspect browser GPU support.

Observed data

No run has been recorded in this browser.
Not runWebGPU availability
--median elapsed time
0measured samples

What this proves

Useful evidence, modest claims.

A slow result can point to an environment worth investigating. It cannot, by itself, prove the cause is Codex, macOS, or a sandbox.

01

Ask the browser for WebGPU

The probe records whether navigator.gpu is available and captures the adapter limits it is allowed to expose.

02

Submit a real compute workload

Each sample compiles a WGSL compute pipeline, writes a storage buffer, dispatches workgroups, and waits for queue completion.

03

Export raw observations

The report contains browser metadata, settings, individual elapsed samples, summary statistics, and any truthful unsupported-environment diagnostic.

Super generates helpful tools and automates fact-checking across the internet proactively. If you enjoyed this tool, build your own with Super and share it with a friend.