August 19, 2025, is undoubtedly a day worth remembering in the history of browser development. Firefox 142 has officially launched, and this is more than just a routine version update; it’s a significant step for Mozilla in bringing AI capabilities into the extension ecosystem. With this version, extension developers can, for the first time, directly call local large language models (LLMs) like Llama from within their add-ons, without relying on remote cloud services.
What’s New in Firefox 142?
Similar to previous version updates, Firefox 142 still includes several minor improvements to the user experience. For instance, dragging and dropping Blob images will be smoother, and the scrolling speed of the bookmark dialog has also been optimized. While these small detail enhancements are minor, they make the daily browsing experience more natural and fluid.
In terms of developer tools, this update introduces two practical APIs: the Prioritized Task Scheduling API allows developers more flexibility in controlling task priority, and the URLPattern API provides a unified way to match and process URLs. These tool enhancements lay the foundation for frontend performance tuning and the implementation of more complex application logic.
The biggest highlight of this version is the opening up of local AI capabilities. Mozilla has introduced a new wllama API in Firefox 142. It is based on WebAssembly and can directly call runtime frameworks like Llama.cpp. This means extension developers can load and run large language models in a completely local environment, and users’ data will not need to leave their personal computers.
More importantly, this feature is very easy to enable. Just type about:config into the address bar and search for “wllama” to enable or disable the related capabilities.
Why Is This a Boon for Developers?
From a developer’s perspective, the arrival of local AI capabilities brings three major advantages:
First is offline functionality. In the past, many AI extensions relied on remote servers to call models, and developers had to bear additional API costs. The wllama API allows extensions to perform inference locally, which both saves costs and greatly improves availability.
Second is privacy. Globally, user concerns about data security and privacy are increasing. Firefox 142’s local AI mode means that users’ input text is no longer uploaded to third-party servers. This not only enhances security but also better aligns with compliance requirements.
Finally, there is creative space. Developers can build entirely new extensions around wllama, such as a “local writing assistant,” an “offline translation tool,” or even a “privacy-first knowledge Q&A plugin.” The official Bugzilla has already set up a dedicated “Enable wllama for extensions” section and is regularly releasing sample code and performance optimization guides to help developers get started quickly.
Conclusion: Firefox 142 Opens a New Channel Between Browsers and AI
The Firefox 142 release date is set for August 19, 2025, but its significance goes far beyond a single point in time. This version enables browser extensions to call local AI large models for the first time, giving users stronger privacy protection and developers greater freedom and creativity.
It’s foreseeable that with the widespread adoption of the wllama API, the future browser extension ecosystem will see an emergence of more innovative, offline intelligent plugins. For users and developers who care about AI, Firefox 142 may be the true starting point for the next generation of browsing experiences.
Version：developer.mozilla.org/en-US/docs/Mozilla/Firefox/Releases/142
