WebGPU Browser AI Test
Check whether your browser can run local LLMs, JAX, and WebGPU AI tools before downloading a model.
--
out of 100
Testing local AI fit
Benchmark in progressThe profiler is checking your GPU adapter, memory signals, and WebGPU compute path.
Confidence: Pending. This test stays on your device and does not upload prompts, hardware details, or benchmark results.
WebGPU support can vary by browser, operating system, GPU driver, and privacy settings.
The score combines adapter limits, local compute timing, CPU threads, and available memory signals.
Checking
Checking
Checking
Checking
Local LLM chat
UnavailableBest first download: compact model (small).
JAX numerical workspace
UnavailableMatrix math, gradients, and notebook-style experiments should run locally in this browser.
Text tools
Start smallCompact text tasks are the safest starting point on this setup.
Vision and multimodal demos
Start smallPrefer small image models and short sessions until you know the browser stays responsive.
- Browser
- Checking
- Platform
- Checking
- CPU threads
- Unknown
- Device memory signal
- Private or unavailable
- Storage quota signal
- Unknown
- GPU adapter
- Checking
- Vendor
- Unknown
- Architecture
- Unknown
- Max buffer size
- Unknown
- Compute workgroup storage
- Unknown
Feature detection is running.
The benchmark runs locally through the browser WebGPU API. It does not send hardware details or test results to WebGPU.ai.
Open JAX WorkspaceWhat is WebGPU?
WebGPU lets websites use modern GPU hardware for compute and graphics, which helps local tools stay responsive.
Can my browser run local chat?
If WebGPU is available and the score is healthy, start with the recommended compact model before trying larger downloads.
Is this a GPU benchmark?
It is a lightweight readiness check, not a full hardware benchmark or a guarantee of model speed.