WebGPU is the browser capability that lets on-device AI run on the visitor's own graphics chip. This tool detects it in your browser, names your GPU adapter, reads the key limits and FP16 support, and estimates which model sizes are realistic on your hardware — plus a WebNN check. It runs entirely in your tab. Background: WebGPU shipped everywhere, and the In-Browser LLM Fit Calculator.
The scan runs automatically on load, entirely in your browser. It requests a high-performance GPU adapter and reads what the browser is willing to tell a web page. Nothing is sent anywhere.
"Supports WebGPU" and "can run the model you had in mind" are two different questions. A browser can expose WebGPU while the actual adapter is too memory-limited for anything but the smallest models, and the number that tells you which is buried in the adapter's limits. This tool surfaces the ones that matter for AI — the biggest buffer it will allocate, whether it supports 16-bit floats, and the compute limits — and turns them into a plain-language read on model size. The estimate is a planning aid, not a benchmark.