The Web Browser has explicit hardware platform requirements, and its AI functionality relies exclusively on the specialized computing architecture of the Apple Silicon chips. This is determined by the underlying technical solution: the project uses the Apple MLX framework to directly call the neural engine (16-core NPU) of the M1/M2 series chips to accelerate the inference operations of the Gemma model.
In terms of technical details, macOS 14.0 and above provide a unified machine learning interface through the MLX framework, which requires the chip to have at least 8 cores of GPU and 16 cores of neural engine in order to ensure the real-time response of the AI sidebar. Test data shows that the latency of generating web page summaries on the M1 Pro chip is controlled within 2 seconds, while traditional Intel chips take more than 15 seconds to process the same task due to the lack of a dedicated AI acceleration unit.
The installation guide specifically states that developers must build the project using Xcode 15.0+, which is configured with compute shaders optimized for Apple Silicon and Metal performance shaders. The pre-compiled version (Web-0.0.2.dmg), which users can download from the GitHub release page, already includes the arm64 native binaries.
This hardware dependency is both a technical limitation and an advantage: while limiting the user base to Apple Silicon device holders, it ensures optimal performance and energy-efficiency ratios for AI functionality. Compared to the cross-platform Electron solution, this native implementation gives Web a 10x+ advantage in energy efficiency ratio.
This answer comes from the articleWeb - macOS AI Browser: a native AI-powered browser for macOSThe
English 
简体中文 
日本語 
Deutsch 
Português do Brasil