{"id":84386,"date":"2026-03-12T10:57:54","date_gmt":"2026-03-12T17:57:54","guid":{"rendered":"https:\/\/blog.mozilla.org\/?p=84386"},"modified":"2026-03-12T11:18:07","modified_gmt":"2026-03-12T18:18:07","slug":"ai-powering-firefox-shake-to-summarize","status":"publish","type":"post","link":"https:\/\/blog.mozilla.org\/en\/firefox\/firefox-ai\/ai-powering-firefox-shake-to-summarize\/","title":{"rendered":"Under the hood: The AI powering Firefox\u2019s Shake to Summarize"},"content":{"rendered":"\n
\"Three<\/figure>\n\n\n\n

We recently released a feature in the Firefox iOS mobile app called \u201cShake to Summarize<\/a>\u201d. The reception was remarkably positive, earning an honorable mention on Time Magazine\u2019s best inventions of 2025<\/a>. <\/p>\n\n\n\n

For anyone unfamiliar with Shake to Summarize, it\u2019s just what the name implies: when you\u2019re browsing a webpage, you can shake your phone to generate a short summary of the page\u2019s content. <\/p>\n\n\n\n

The gesture is fun, the feature is useful, and the whole thing feels simple and natural.<\/p>\n\n\n\n

From a technical standpoint, the application works just how you\u2019d imagine: when a shake (or lightning bolt-icon press) is detected, we grab the web page content, pass it to an LLM for summarization, and then return the result to the user. <\/p>\n\n\n\n

But with the LLM landscape being as vast as it is, there is a lot to consider when bringing even a relatively straightforward application to the market. In this post, we\u2019ll discuss the ins and outs of our approach to model selection. We will leave prompt development and quality testing for a future article.<\/p>\n\n\n\n

Which model? <\/h2>\n\n\n\n

These days, there are many LLMs available, with a steady stream of new releases arriving almost every week.. Each release is paired with a slate of benchmark scores, showing the new model\u2019s superiority along one dimension or another. The pace of development has been fast and furious and billions of dollars have been spent inching the numbers higher and higher.

But what do these metrics mean in practice? At the end of the day, we are building a product for users. The most important metric for us is, \u201chow useful are the summaries the model produces?\u201d – something that isn\u2019t neatly captured by the benchmark scores. To select the best model for our applications, we need to run our own tests.<\/p>\n\n\n\n

For us, the best model had to excel along several dimensions: <\/p>\n\n\n\n