WebXPRT FAQ

We rely on web-connected devices to check movie times, figure out where to eat, and communicate with our friends and family. But there are thousands of devices to choose from, all claiming that they have great web browsing capabilities. How can you tell how good these devices really are?

WebXPRT is the answer: a free online tool that gives you objective information about how well a laptop, tablet, smartphone, or any other web-enabled device handles common web tasks. By running a simple test on your device at WebXPRT.com, you can get a score that makes it easy to compare your device against others on the WebXPRT website. Or just check out results from a variety of laptops, tablets, and phones at WebXPRT.com.

WebXPRT is one of the BenchmarkXPRT benchmarks, a set of free, easy-to-use tools that help you learn how all kinds of devices perform.

A benchmark, in this case, is a tool designed to test a wide range of devices on a level playing field. When you look at results from BenchmarkXPRT benchmarks, you get unbiased, fair product comparisons.

Technology manufacturers, developers, and sellers; tech-review websites like Notebookcheck.net and PCMag; and consumers like you. WebXPRT is a great resource for anyone who wants to compare devices against each other on the battlefield of web browsing.

WebXPRT tells you how well any connected device handles both common and advanced web-browsing tools. To do this, WebXPRT runs seven online tests that simulate the kinds of things you do online every day. It measures how long (in milliseconds) your device takes to complete each of those tasks. Then it uses those times to calculate a single-number overall score, which you can use to compare your device against others at WebXPRT.com.

WebXPRT 5 also offers an optional experimental AI workload for high-end systems: Image Generation. The Image Generation workload runs completely separately from WebXPRT 5’s main tests and produces a single score, measured in images per minute. The main WebXPRT 5 testing experience and scores will not change when you run the experimental workload.

It runs these seven tests seven times each:

1. Video Background Blur with AI: Blurs the background of a video call using an AI-powered Segmentation model.
2. Detect Faces with AI: Detects faces and organizes photos in an album using computer vision (OpenCV.js with Caffe Model).
3. Image Classification with AI: Labels images in an album using machine learning (OpenCV.js and ML Classify with the SqueezeNet model).
4. Document Scan with AI: Scans a document image and converts it to text using ML-based OCR (Wasm with LSTM).
5. Photo Effects: Applies a filter to six photos using the Canvas API.
6. School Science Project: Processes a DNA sequencing task using Regex and String manipulation.
7. Homework Spellcheck: Spellchecks a document using Typo.js and Web Workers.

The tests use different combinations of HTML5, Canvas, JavaScript, and WebAssembly—common elements in many web pages—to gauge how well your device and browser work together in everyday web browsing situations.

WebXPRT 5 also offers an optional experimental AI workload for high-end systems: Image Generation. You can read more about that test in the sections about the experimental workload below.

It’s easy to run the main tests – no download required!

1. You don’t want the device to shut down during the tests, so go to Settings and set the sleep time to at least half an hour. This should cover the time it takes to run all the tests.
2. Close all your other apps and windows. Other activity on your device might lower your score.
3. Go to WebXPRT.com and click "Run WebXPRT 5." Click “Continue” in the popup box.
4. WebXPRT will run all seven core workloads by default. If you only wish to run selected workloads, uncheck the box next to each workload you don’t wish to run. However, to get an Overall score, you will need to run all seven workloads.
5. Click the Start test button at the bottom of the page, and WebXPRT will begin!

WebXPRT will run each of its seven core tests seven times. The progress bar at the top of the screen lets you keep track of the benchmark’s progress. WebXPRT also displays the ratio of completed workloads and completion percentage in the upper right-hand corner. While it’s running, try not to touch or click the screen; this won’t hurt your device, but it might affect the results. When you see the words "Test Complete!" followed by the score information, the test is finished. The score under Your Score is all you need to compare your device against others at WebXPRT.com. If you’re interested, however, you can also scroll down to see how fast your device completed each individual task.

WebXPRT’s optional experimental Image Generation workload uses on-device text-to-image AI technology to generate images from text prompts using Z-Image-Turbo with ONNX Runtime Web and WebGPU. The Image Generation workload runs completely separately from WebXPRT 5’s main tests and produces a single score, measured in images per minute. The main WebXPRT 5 testing experience and scores do not change when you run the experimental workload.

It’s easy to run the experimental Image Generation workload, but there are a few points to keep in mind:

• The workload is completely optional, and it runs separately from WebXPRT 5’s main tests. Your main WebXPRT 5 scores and testing experience will not change.
• It won’t run on everything. WebXPRT’s main tests run on just about anything with a browser. Because the Image Generation workload is a cutting-edge, resource-intensive, experimental performance test, it won’t cover that same broad range of devices. It will best suit late-model, high-performance laptops and desktops. We recommend testing on systems with at least 32GB RAM and a minimum of 8 CPU cores.
• It will require a large one-time model download. The first time you run the Image Generation workload on a specific system, WebXPRT will download the Z-Image-Turbo model from the BenchmarkXPRT Hugging Face repository. The total size of the model download is just under 6.5 GB, so depending on your internet speed, the download can take 5 to 20 minutes.
• Expect longer runtimes than a typical WebXPRT 5 test. After the model download, each run may take about five minutes.

When you’re ready to run the Image Generation workload, follow these steps:

1. You don’t want the device to shut down during the tests, so go to Settings and set the sleep time to at least half an hour. This should cover the time it takes to download the model and run the tests.
2. Close all your other apps and windows. Other activity on your device might lower your score.
3. Go to WebXPRT.com and click "Run WebXPRT 5." Click “Continue” in the popup box.
4. Click the “Experimental AI workloads” button below WebXPRT 5’s normal “Start test” button.
5. On the Experimental AI workloads page, make sure that the checkbox for the Image Generation workload is checked, and click “Start test.”

WebXPRT will then initiate the model download. Once the download is complete, the workload will automatically produce four prompt-based images: an initial warm-up image and three successive scored images.

If the Image Generation workload successfully completes, it will report a single score, measured in images per minute. Due to the nature of GenAI model technology, the workload unavoidably generates slightly different image outputs each time, so no two runs look exactly alike. That said, the outputs are stable enough to produce consistent scores with low variance.

Your Score:

• This overall score is all you need to check your results against the results of other devices on WebXPRT.com. Here, a higher number is better.
• In the test details, WebXPRT also generates a Test ID and time stamp each time the program runs and reports this information along with the browser version.
• Below the Test ID, the reported +/- number is the confidence interval.

Individual Test Results:

• Each test lists its time in milliseconds (ms). Here, a lower number is better.
• The +/- following the (ms) time is the confidence interval.
• The confidence interval shows an estimated variation in the overall WebXPRT score based on a 95% confidence interval calculated from multiple test iterations. So, for example, if your score was 195 and your confidence interval was 2, there is a 95% likelihood that the next test will produce a score between 193 and 197. A smaller number is better, because it means that next time you run the test, it’s more likely that you’ll get the same score.

Yes. WebXPRT tests the capabilities of your device/browser combination. We recommend you run the test on the browser you use most often. That said, WebXPRT is also a great way to test the speed of several different browsers and see which browser/device combination is fastest.

No. It may cause the test to take longer to complete, but WebXPRT doesn’t need a perfect Internet connection.

On the Results page, click the “Submit your results” button.

1. Fill in your email and the name of your device.
2. Fill in the model number. This is in a different place for each device, but you should be able to find it by searching “Model number for [device name].” If your device does not have a model number, put N/A.
3. Fill in the processor. Please include the processor vendor name and model number (e.g. Qualcomm Snapdragon 888 or Intel Core i9-12900HK).
4. Fill in the memory (e.g. 12 GB).
5. Fill in the operating system.
6. Fill in the browser and version.
7. Fill in any additional information you would like to include.
8. Click “Submit”, and you’re done!

We vet all submissions before we publish them on the website, so you may not see your results for a few days. If there’s a problem with the results, we’ll contact you by e-mail.

Go to WebXPRT.com, and click View Results.

Go to WebXPRT.com or follow us on Twitter and Facebook. We announce breaking news on the BenchmarkXPRT blog (available to everyone) and the BenchmarkXPRT forums (available to members only). If you can't find the answer to your question, or you need help with WebXPRT, send an email to our team at [email protected].