Google just shipped a native Gemini app for macOS. This isn't another Electron wrapper hogging your RAM. The app is a proper ARM64 binary built for Apple Silicon, requiring macOS Sequoia (15.0) or later. Intel Mac owners are out of luck. You summon it with Option+Space from anywhere, pulling up a mini-chat overlay without switching to a browser tab. This represents the debut of the native Gemini app. Chat history and memory sync across desktop, web, and mobile if you're signed into the same Google Account.
Screen sharing is the standout feature. Share a specific window with Gemini and it reads what's visible to give contextual answers about your documents or code. Grant Accessibility permissions and it can parse full browser pages. The app also bundles image generation tools through Nano Banana, video generation with Veo, Gemini Live conversations, and Deep Research. That's a lot of capability sitting in your menu bar.
Going native isn't cosmetic. By targeting ARM64 directly, Google gets access to the Apple Neural Engine for hardware-accelerated ML tasks and can exploit the Unified Memory Architecture of M-series chips for faster data movement between CPU, GPU, and Neural Engine during inference. This matters for low-latency features like real-time screen analysis. SwiftUI likely powers the interface, and the app uses native macOS APIs for system permissions like screen recording and input monitoring rather than web workarounds.
Community feedback so far splits along familiar lines. Fans of the omnipresent ask bar say it finally makes Gemini feel like part of the OS. Skeptics ask a fair question: if the web app works fine, why install another thing? Some are already hoping for CLI integration, which would make this genuinely useful for developers. The app is free, available everywhere Gemini is supported, and needs no subscription. Whether native access beats a pinned browser tab is between you and your menu bar.