BenchmarkXPRT Blog banner

Category: Web-based testing

WebXPRT 5: AI tests now, lots of room for growth

Justin Greene
By Justin Greene
on July 9, 2026

In past blog posts, we’ve discussed our goal of developing one or more experimental WebXPRT workloads focused on local, browser-side AI technologies. While many of us regularly interact with cloud-based AI apps and services through a browser, on-device AI capabilities are growing rapidly, and we want WebXPRT to continue to evolve with them.

There are several driving factors behind that growth. Web API technologies keep maturing, giving browsers direct access to the hardware they need for real inference work. Advanced GPU and NPU technology is now widely available in consumer devices, so the local computing power necessary to run AI applications on-device is in reach for many users. And for many organizations, there are compelling reasons to execute increasingly vital work like LLM inferencing and agentic coding tasks on local machines—such as data privacy, regulatory compliance, and cost control.

The reasons for the experimental workload approach

The expansion of on-device AI is exactly the type of shift we built the experimental workload concept to capture. As we shared when we first announced the WebXPRT 5 workload lineup, an experimental workload section gives us the flexibility to put cutting-edge measurement tools in users’ hands—even if those tools won’t yet run on every platform WebXPRT has traditionally supported. Experimental scores stay separate from the main overall score and are completely optional, so we can add tests without affecting comparability or asking anyone to retest. That approach maintains WebXPRT’s strengths while preparing the benchmark for the future—and giving all of us valuable information today.

The AI functions that WebXPRT 5 measures today

WebXPRT 5 already includes four workloads that utilize AI capabilities: Video Background Blur with AI, Detect Faces with AI, Image Classification with AI, and Document Scan with AI. These workloads use machine learning—computer vision and OCR models such as a Caffe-based face detector, SqueezeNet for image labeling, and an LSTM-based OCR engine. WebXPRT’s ability to measure how well devices handle those types of workloads has real value, and it reflects the kinds of light browser-side inference tasks that have been in widespread use for a while.

We recognize, though, that there’s a clear need for more demanding local, browser-based AI workloads—especially LLM inference. We’re targeting that need with our experimental work. Like pretty much everyone else, we’re also developing in the midst of an incredibly dynamic technical environment. We want to purposefully move forward without sacrificing WebXPRT’s stability and reliability for the sake of expedience.

The main decisions we face

Choosing a Web AI framework. We’re still researching our open-source framework options, including candidates like ONNX Runtime Web, Transformers.js, MediaPipe, and TensorFlow.js. The ground here continues to shift. For example, Transformers.js v4 now supports a WebGPU backend and spans a very broad range of model architectures. So, one of our ongoing challenges is picking a durable foundation.

Choosing a web API. Of the primary options we’re investigating, WebGPU now has the broadest browser support (Chrome, Edge, and partial support in Firefox and Safari). WebNN remains the most promising option in the long term because it can directly target NPUs, but it’s still not ready for production—its W3C spec only reached Candidate Recommendation status in early 2026, and browser support outside of flagged, experimental builds isn’t there yet. Our web API outlook hasn’t changed much from before: WebGPU is the most practical path today, and WebNN may be an exciting possibility for tomorrow.

Choosing and sizing workloads. We’ll ideally find workloads demanding enough to genuinely stress new hardware, but light enough to run on slightly older gear without forcing huge model downloads or overextending the test’s runtime. The sweet spot for browser inference today tends to be small, quantized models, and memory ceilings and cold-start downloads are real constraints. Striking the right balance is another part of the challenge we’re working through.

We appreciate your patience

We’ve been talking about experimental WebXPRT AI workloads for a while. While we wish we already had everything worked out, we think the end product will be worth the wait. We appreciate your patience as we work through the details, and we’ll keep updating you here in the blog as we make progress.

As always, we’re open to suggestions. If you have ideas for a browser-based AI workload scenario, a framework or API you think we should weigh, a browser-based AI application you want us to consider, or any other related thoughts, please let us know!

Justin

Posted in AI, benchmark, Benchmarking, BenchmarkXPRT, browser performance, Browser-based benchmarks, Caffe, Chrome, Chromium, Collaborative benchmark development, computer vision, Cross-platform benchmarks, Future of performance evaluation, GPU, HTML5, image classification, inference, JavaScript, large language models, MediaPipe, Microsoft Edge, NPU, OCR, ONNX Runtime Web, Performance benchmarking, Performance of computing devices, Safari, SqueezeNet, TensorFlow, Wasm, Web AI, web API, Web-based testing, WebAssembly, WebGPU, WebNN, WebXPRT, WebXPRT 5 | Tagged , , , , , , , , , , , , , , , , , , , , , , , , , |

WebXPRT 5 source code access is on the way

Justin Greene
By Justin Greene
on March 12, 2026

Recently, a member of the tech press asked us if we were planning to offer a way for users to set up an offline version of WebXPRT 5 for locally hosted tests. The short answer is “yes.”

The long answer is that the question provides us with a good opportunity to talk about XPRT source code access and let new users know how it works.

Since the early days of the BenchmarkXPRT Development Community, we’ve provided free access to the benchmark source code. We believe that by publishing XPRT code and allowing interested parties to access and review that code, we’re doing our part to encourage transparency and honesty in the benchmarking industry.

While we offer free access to the XPRT source code, our approach to derivative work differs from some traditional open-source models that encourage developers to alter products and even take them in substantially different directions. Because benchmarking requires a product that remains static to enable valid comparisons, we prioritize maintaining the integrity and consistency of the benchmark over time. So, we allow people to download the source, but we also reserve the right to control derivative works. This approach discourages a situation where someone publishes an unauthorized version of the benchmark and calls it an “XPRT.”

For WebXPRT 5, we’ll offer the code in the form of a build package—containing all the necessary files and instructions—that will be available upon request. By offering the code upon request, as opposed to posting a permanent download link, we can prevent bots or other malicious actors from downloading it. This method also lets us engage with users interested in the source code and answer their questions.

With the WebXPRT 5 build package, you’ll be able to set up your own WebXPRT 5 instance for purposes of review, internal testing, or experimentation. We do ask that you publish only test results from the official version of WebXPRT 5 that we host at WebXPRT.com.

We expect to have the build package ready within the next few weeks. When it’s available, we’ll let readers know here in the blog, and we’ll provide more details about the access and setup process.

If you have any questions about accessing the WebXPRT 5 source code, please let us know!

Justin

Posted in benchmark, Benchmarking, BenchmarkXPRT, BenchmarkXPRT development community, browser performance, Browser-based benchmarks, Collaborative benchmark development, Performance benchmarking, Source code, tech press, Web-based testing, WebXPRT, WebXPRT 5 | Tagged , , , , , , , , , |

The WebXPRT 5 Preview is here!

Justin Greene
By Justin Greene
on December 16, 2025

We’re excited to announce that the WebXPRT 5 Preview is now available!

The Preview is available to everyone. You can access it at www.WebXPRT5.com or through a link on WebXPRT.com.

You are free to publish scores from testing with this Preview build; in fact, we encourage it. We want to know how it is performing for you, so we love to see both test scores and any feedback you would like to give.

We may still tweak a few things in the benchmark between this Preview and the final release. We plan to limit any potential changes, however, to areas like the UI and other features, things we do not expect to affect test scores.

Longtime WebXPRT users will notice that while the WebXPRT 5 Preview UI has a new look and feel, the basic layout has not changed. The general process for kicking off both manual and automated tests is the same as with WebXPRT 4, so the transition from WebXPRT 4 to WebXPRT 5 testing should be straightforward.

We also encourage you to check out our recent XPRT blog post on the WebXPRT 5 workload lineup for more details about what’s new in the Preview release—including more AI-oriented scenarios than ever before!

After you try the WebXPRT 5 Preview, please send us your comments. Thanks, and happy testing!

Justin

Posted in AI, Automation, benchmark, Benchmarking, BenchmarkXPRT, browser performance, Browser-based benchmarks, Collaborative benchmark development, Future of performance evaluation, Performance benchmarking, Web-based testing, WebXPRT, WebXPRT 4, WebXPRT 5 | Tagged , , , , , , , , , , |

WebXPRT 5: The workload lineup

Justin Greene
By Justin Greene
on December 8, 2025

The WebXPRT 5 development process heading into the final stretch, so we’d like to share more information about the workloads you’re likely to see in the WebXPRT 5 Preview release—and when that release may be available. We’re still actively testing candidate builds, studying results from multiple system tests, and so on, so some details could change. That said, we’re now close enough to provide a clearer picture of the workload lineup.

Core workloads

WebXPRT 5 will likely include the following seven workloads:  

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

The sub-scores for each of these tests will contribute to WebXPRT 5’s main overall score. (We’ll discuss scoring in future blogs.)

Experimental workloads

We’re currently planning to include an experimental workload section, something we’ve long discussed, in WebXPRT 5. Workloads in this section will use cutting-edge browser technologies that may not be compatible with the same broad range of platforms and devices as the technologies in WebXPRT 5’s core workloads. For that reason, we will not include the scores from the experimental section—in the Preview build and future releases—in WebXPRT 5’s main overall score.

In addition, WebXPRT 5’s experimental workloads will be completely optional.

Moving forward, WebXPRT’s experimental workload section will provide users with a straightforward way to learn how well certain browsers or systems handle new browser-based technologies (e.g., new web apps or AI capabilities). We’ll benefit from the ability to offer workloads for large-scale testing and user feedback before committing to including them as core WebXPRT workloads. Because future experimental workloads will run independently of the main test, we can add them without affecting the main WebXPRT score or requiring users to repeat testing to obtain comparable scores. We think it will be a win-win scenario in many respects.  

We’re still evaluating whether we can finish the first experimental workload in time to include it in the WebXPRT 5 Preview release, but we will definitely have at least the section and the framework for adding such a workload. When we are confident that an experimental workload is ready to go, we’ll share more information here in the blog and be all set up to incorporate it.

Timeline

If all goes well, we hope to publish the WebXPRT 5 Preview very soon, followed by a general release in early 2026. If that timeline changes significantly, we’ll provide an update here in the blog as soon as possible.

What about an “AI score”?

We’re still discussing the concept of a stand-alone WebXPRT 5 “AI score,” and we go back and forth on it. That score would combine WebXPRT’s AI-related subscores into a single score for use in AI capability comparisons. Because we’re just now beefing up WebXPRT’s AI capabilities, we’ve definitely decided not to include an AI score right now. We would love your feedback on the concept as we plan WebXPRT’s future. If that’s something that you would be interested in, please let us know!

If you have any questions about the WebXPRT 5 details we’ve shared above, please feel free to ask!

Justin

Posted in AI, benchmark, BenchmarkXPRT, BenchmarkXPRT development community, browser performance, Browser-based benchmarks, Caffe, Collaborative benchmark development, Cross-platform benchmarks, Future of performance evaluation, image classification, image processing, JavaScript, LAMP, on-device AI, Performance benchmarking, Performance of computing devices, SqueezeNet, Wasm, Web AI, web API, Web-based testing, WebXPRT, WebXPRT 4, WebXPRT 5 | Tagged , , , , , , , , , , , , , , , , , , , , , , |

You asked, and we heard you: WebXPRT 5 is on the way!

Justin Greene
By Justin Greene
on October 9, 2025

We’re excited to announce that WebXPRT 5 is officially on the way! Since we launched WebXPRT 4 in February 2022, it’s proven to be an exceptionally successful and reliable go-to benchmark for OEM labs, the tech press, and individual users alike—to the tune of over 644,000 runs to date. In past blog posts, we’ve discussed new features and possible auxiliary workloads that we contemplated adding to WebXPRT 4. As we’ve considered user comments and suggestions, changes in web technology, and how we can best position WebXPRT as a relevant browser benchmark in the future, however, it became clear that it was time for an all-new WebXPRT.

Now that we’ve announced WebXPRT 5, the first question for many existing WebXPRT users may be, “When will WebXPRT 5 be available?” We’re not yet ready to share an anticipated WebXPRT 5 release date, but we can share that a lot of groundwork is already complete, and the remaining work is moving along rapidly. We’ll continue to issue updates here in the blog as we reach important milestones.

The second question for many existing WebXPRT users may be, “How will WebXPRT change?” We’re not yet ready to share extensive details about WebXPRT 5’s workloads—rest assured that we will as soon as we can firm up everything—but we can share a few key guidelines we tried to follow in our WebXPRT 5 design. Each of these points of emphasis is a result of feedback we’ve received from labs, as well as features that users have asked for.

  • Provide more AI-related workloads. In past blog posts, we’ve discussed the growing importance of local, browser-side AI. WebXPRT 4 already includes timed AI tasks in two of its workloads: the Organize Album using AI workload and the Encrypt Notes and OCR Scan workload. We’re working on ways to expand WebXPRT’s AI portfolio in the next version.
  • Add WebGPU workloads. As a web API, WebGPU enables web-based applications—such as image-based GenAI and inference workloads—to directly access the graphics rendering and computational capabilities of a system’s GPU. We hope to incorporate WebGPU measures in WebXPRT 5.
  • Improve WebXPRT’s utility as a tool for test labs, publications, and engineering analysis.
    • Update the workloads with longer operations. Many of WebXPRT’s existing workloads no longer challenge cutting-edge consumer hardware as much as many of us would like. Testing labs have asked for longer and more demanding workloads. We’re working on incorporating workloads that are accessible enough to be run by a broad range of devices yet challenging enough to allow performance differentiation among high-end systems.
    • Enable more precise performance measures. Labs and testers have also asked for more granular insight into the workloads to help with engineering-level performance analysis. Currently, some WebXPRT 4 workload scores include multiple timed tasks. If we separate those compound scores so that each workload reports results from only one timed task, users will be able to more precisely assess how well a device performs while handling specific operations. We’re looking into this approach.
  • Modernize the harness to make it more flexible and to speed future work. WebXPRT 4’s current harness works with server-side sessions on a LAMP (Linux, Apache, MySQL and PHP) stack. If we implement the harness via JavaScript on the client side, it will pave the way for faster development and testing cycles in the future.

We expect WebXPRT 5 to carry on the WebXPRT legacy of reliability and real-world relevance, while providing users with compelling new workloads and features. As has been our habit with new benchmark releases, however, we won’t force anyone to change versions anytime soon. Instead, we will continue to make WebXPRT 4 available for quite some time after WebXPRT 5 goes live.

If you have any questions or comments about WebXPRT, please let us know!

Justin

Posted in AI, Application-based benchmarks, benchmark, Benchmark metrics, Benchmarking, BenchmarkXPRT, BenchmarkXPRT development community, browser performance, Browser-based benchmarks, Cross-platform benchmarks, Future of performance evaluation, GPU, image classification, image processing, inference, LAMP, on-device AI, Performance benchmarking, Web AI, web API, Web-based testing, WebGPU, WebXPRT, WebXPRT 3, WebXPRT 4 |

Browser-based AI tests in WebXPRT 4: optical character recognition

Justin Greene
By Justin Greene
on July 3, 2025

In our previous blog post, we discussed the rapidly expanding influence of AI-enhanced technologies in areas like everyday browser activity—and the growing need for objective performance data that can help us understand how well new consumer devices will handle AI tasks. We noted that WebXPRT 4 already includes timed AI tasks in two of its workloads—the “Organize Album using AI” and “Encrypt Notes and OCR Scan”—and we provided some technical details for the Organize Album workload. In today’s post, we’ll focus on the Encrypt Notes workload.

The Encrypt Notes workload includes two separate scenarios that reflect common web-based productivity app tasks. The first scenario syncs a set of encrypted notes, and the second scenario uses AI-based optical character recognition (OCR) to scan a receipt, extract data, and then add that data to an expense report.

Here are the details for each scenario:

  • The encrypt notes scenario downloads a set of notes, encrypts that data, temporarily stores it in the browser’s localStorage object (the localStorageDB.js database layer), and then decrypts and renders it for display. This scenario measures HTML5 Local Storage, JavaScript, AES encryption, and WebAssembly (Wasm) performance. 
  • The OCR scan scenario uses a Wasm-based version of Tesseract.js (tesseract-core.wasm.js v2.20) to scan an expense receipt. Tesseract.js is a JavaScript port of the Tesseract OCR engine—a popular open-source C/C++ library that extracts text from images and PDFs. The scenario then adds the receipt to an expense report. This scenario measures HTML5 Local Storage, JavaScript, and Wasm performance. 

We mention this test under the AI umbrella in part because people sometimes use the term “OCR” to refer to a spectrum of AI and non-AI technologies. In this case, though, the specifics make this workload clearly have an AI component. The Wasm-based Tesseract library that we use in WebXPRT 4 is based on a version of C/C++ (v4.x) that uses Long Short-Term Memory (LSTM). LSTM is a type of recurrent neural network (RNN) that is particularly well-suited for processing and predicting sequential data. As such, it is clearly an AI component of the Encrypt Notes and OCR Scan workload.

To produce a score for each iteration of the workload, WebXPRT calculates the total time that it takes for a system to sync (encrypt, decrypt, and render) the notes, use OCR to scan the receipt, and add the scanned data to an expense report. In a standard test, WebXPRT runs seven iterations of the entire six-workload performance suite before calculating an overall test score. You can find out more about the WebXPRT results calculation process here.

Along with our post on the Organize Album workload, we hope this information provides a deeper understanding of WebXPRT 4’s AI-equipped workloads. As we mentioned last time, if you want to explore the structure of these workloads in more detail, you can check out previous blog posts for information about how to access and use the WebXPRT 4 source code for free. You can also read more about WebXPRT’s overall structure and other workloads in the Exploring WebXPRT 4 white paper.

If you have any questions about WebXPRT 4, please let us know!

Justin

Posted in AI, benchmark, Benchmark metrics, Benchmarking, BenchmarkXPRT, browser performance, Browser-based benchmarks, computer vision, Cross-platform benchmarks, HTML5, inference, OCR, on-device AI, Performance benchmarking, Wasm, Web-based testing, WebAssembly, WebXPRT, WebXPRT 4, White papers | Tagged , , , , , , , , , , , , , , , , |

Check out the other XPRTs:

Forgot your password?
© Principled Technologies, Inc. 2026   |   Privacy policy

Cookie Notice: Our website uses cookies to deliver a smooth experience by storing logins and saving user information. By continuing to use our site, you agree with our usage of cookies as our privacy policy outlines.

x