Can I run cross-browser tests without a cloud grid?

Yes, for the browsers Playwright ships. Chromium, Firefox, WebKit and phone emulation run on your own machine, in parallel, from one config file. What a local run cannot give you is Safari itself on a Mac or an iPhone, or physical devices you do not own. The example below ran four browser projects in under 14 seconds on a Windows 11 laptop.

By the Nexus Studio team, Verdict System LLC. October 27, 2026.

What does a cloud grid give you?

Three things: browsers and operating systems you do not have, physical devices, and parallel capacity beyond your own machine. The first one, the same test in several browser engines, you can have locally.

Can I try it on a runnable example?

Save this as playwright.config.js. It declares four projects and lets three workers run them side by side.

import { devices } from '@playwright/test';

export default {
    workers: 3,
    reporter: 'list',
    projects: [
        { name: 'chromium', use: devices['Desktop Chrome'] },
        { name: 'firefox', use: devices['Desktop Firefox'] },
        { name: 'webkit', use: devices['Desktop Safari'] },
        { name: 'phone', use: devices['Pixel 7'] },
    ],
};

Save this as grid.spec.js. It needs no network: it renders a page from a string.

import { test, expect } from '@playwright/test';

test('the page renders in this browser', async ({ page, browserName }) => {
    await page.setContent('<h1>Hello</h1><p id="ua"></p><script>ua.textContent = navigator.userAgent</script>');
    await expect(page.locator('h1')).toHaveText('Hello');
    console.log(browserName, await page.locator('#ua').textContent());
});

Install the browsers once, then run:

npx playwright install
npx playwright test

The run printed this. The lines come out in the order the workers finish, not in the order of the config:

Running 4 tests using 3 workers
  ok 1 [chromium] › grid.spec.js:3:1 › the page renders in this browser (1.6s)
  ok 3 [webkit] › grid.spec.js:3:1 › the page renders in this browser (2.3s)
  ok 4 [phone] › grid.spec.js:3:1 › the page renders in this browser (328ms)
  ok 2 [firefox] › grid.spec.js:3:1 › the page renders in this browser (3.0s)

  4 passed (8.6s)

Three runs took 8.6, 10.6 and 13.5 seconds on an AMD Ryzen 5 7520U laptop with Windows 11. Yours will differ with its processor and memory.

What can I not run on my own machine?

• Real Safari. Playwright's WebKit is a build of the engine, not the Safari app, on Windows and on a Mac. A bug that only the real Safari on macOS or iOS shows needs that browser.
• iPhones and iPads. Phone emulation sets the screen, the user agent and touch, and it is not the device. Nexus Studio runs real Android phones connected to your PC by USB. It does not run iPhones.
• A matrix of old versions. You get the browser versions Playwright installs, not every version of every operating system.

If those matter to you, a cloud grid or a device lab is the right tool for that part, and the comparison with BrowserStack and Sauce Labs says where each fits.

Where does Nexus Studio fit?

Nexus Studio is a Windows 11 desktop app that runs your Playwright, Selenium and pytest tests on your own machines. It adds the recorder, scheduled runs, 14 verdicts and the evidence of each run, and the evidence stays on your disk. It is priced per machine, with a 30-day free trial. It does not replace a cloud grid when you need physical iPhones. The self-hosted Playwright guide covers running the same tests in Docker or on a Linux server.

What are the limits?

• The timing comes from one laptop, and the example touches no network. A real site adds its own loading time.
• The first npx playwright install downloads the browsers, about 1.2 GB for the version Nexus Studio ships, so that run needs internet.
• Each browser needs about 0.5 GB of free memory by the engine's own estimate, so the number of workers is bounded by your RAM, and Nexus Studio asks for 16 GB as the minimum.

Where can I see it?

The own-your-infrastructure page says what stays on your machines and what a hosted grid would cost you, and the getting started guide lists the requirements. The 30-day trial runs the same tests with verdicts and evidence.