Hardware and system requirements

Chrome's built-in AI APIs (Prompt, Summarizer, Writer, Rewriter) require Gemini Nano, which has minimum hardware thresholds. Edge shares these requirements for Chromium-based APIs. Firefox AI features do not currently use on-device models.

Source: Chrome built-in AI — Get started (May 2025). Edge inherits these Chromium requirements.

Chrome: enable built-in AI

Step-by-step for Chrome Stable, Beta, or Canary.

  1. Update Chrome to the latest version (138+ recommended). For earliest access, install Chrome Canary.
  2. Navigate to chrome://flags/#optimization-guide-on-device-model and set it to Enabled.
  3. Navigate to chrome://flags/#prompt-api-for-gemini-nano and set it to Enabled (or Enabled multilingual for non-English support).
  4. Click Relaunch at the bottom of the flags page.
  5. After relaunch, open chrome://on-device-internals and check the Model Status tab. Wait for the model download to complete (this can take several minutes on first run).
  6. Open the demo page, click Run Built-in Tests, and confirm the Prompt API shows "Detected and ready".
Troubleshooting: If the model shows "downloadable" but never downloads, confirm your device meets the hardware requirements above, restart Chrome, then revisit chrome://on-device-internals. You can also run await LanguageModel.availability() in the DevTools console — it should return "available". See full docs.

Edge: enable built-in AI

Edge shares Chromium's built-in AI APIs alongside its own Copilot features.

  1. Update Edge to the latest version (138+ recommended). For early access, install Edge Dev or Canary.
  2. Navigate to edge://flags/#optimization-guide-on-device-model and set it to Enabled.
  3. Navigate to edge://flags/#prompt-api-for-gemini-nano and set it to Enabled.
  4. Click Restart at the bottom of the flags page.
  5. After restart, open edge://on-device-internals and check the Model Status tab. Wait for the model download to complete.
  6. Open the demo page, click Run Built-in Tests, and confirm the Prompt API shows "Detected and ready".
Note: Edge also includes Copilot, which is a cloud-based AI assistant. Copilot features work independently of the on-device model flags above. The built-in AI APIs tested by this demo are the on-device Chromium APIs, not Copilot.

Firefox: AI controls

Firefox takes a different approach — AI features emphasize user control and privacy.

  1. Install Firefox Beta or Nightly for newest features.
  2. Open about:support to verify your channel and build version.
  3. Open Firefox Settings → Firefox Labs and review available AI experiments.
  4. Enable any available AI chatbot integrations or sidebar features.
  5. Open the demo page, click Run Built-in Tests to see which APIs are detected.
Firefox does not currently expose the same on-device AI APIs as Chrome/Edge (Prompt API, Summarizer, etc.). The built-in test panel will show "Not detected" for most APIs. Firefox AI features are currently focused on sidebar/chatbot integrations rather than on-device inference APIs.

Safari: current status

Safari and WebKit do not yet expose built-in AI APIs.

Apple has not announced plans for on-device AI APIs in Safari comparable to Chrome's Prompt API. Apple Intelligence features are integrated at the OS level rather than through browser JavaScript APIs.

The built-in test panel will show "Not detected" for all AI APIs in Safari. This is expected. Apple may introduce Safari-specific AI APIs in the future.
Browser Track Release Speed AI Experiment Access Best Use
Chrome Stable Slowest Limited Baseline and production comparison
Chrome Beta Fast Good Pre-release demos with moderate stability
Chrome Canary Fastest Highest Earliest AI capability exploration
Edge Stable Slowest Limited Baseline with Copilot integration
Edge Dev / Canary Fastest Good On-device API testing alongside Copilot
Firefox Release Slowest Limited General audience baseline
Firefox Nightly Fastest Highest Early feature and experiment validation

How to show this to people during a demo

  1. Open this page first and walk through channel setup.
  2. Open the main demo in each installed browser channel.
  3. Run the built-in test panel and compare results side by side.
  4. Capture screenshots of each channel so differences are visible in reports.