Google started handing out early builds of a dedicated Gemini app for macOS this week, according to Bloomberg's Mark Gurman. The app, internally codenamed Janus, is being distributed to participants in Google's consumer beta program. Both OpenAI and Anthropic have had native Mac apps for their respective chatbots for some time now, which makes this less of a bold move and more of a gap being plugged.
Until this week, the only way to use Gemini on a Mac was through a browser tab. That's it. No keyboard shortcuts, no local file integration, no quick-invoke from anywhere in the OS. For a company spending billions on AI infrastructure, the absence of a Mac client has been a strange omission.
Desktop Intelligence is the interesting part
The beta itself looks a lot like the existing Gemini apps on iPhone and iPad, per details shared by app researcher M1Astra. You can generate images, video, music, charts. You can search the web, upload documents, keep a conversation history. Standard chatbot stuff.
But buried in the app's code is a feature called Desktop Intelligence, and this is where things get more interesting. A string found in the code reads: "When you enable apps for Desktop Intelligence you are enabling Gemini to see what you see (such as screen context) and pull content directly from these apps to improve and personalize your experience only when Gemini is in use."
Screen context. Pull content from apps. That's the pitch, anyway.
Claude's desktop app already does something similar through its Cowork feature, and ChatGPT's Mac app can read your screen too. So Google isn't inventing a new category here. It is catching up to one. The question is whether Desktop Intelligence will go beyond reading and actually take actions inside apps, the way Cowork does. Engadget notes that Google already offers limited agentic capabilities on smartphones, so desktop actions aren't out of the question. But Google's spokesperson declined to comment on any of it, which tells you roughly how far along they are.
Why now?
The timing makes more sense when you remember what happened in January. Apple and Google announced a partnership on January 12 under which Gemini models would power the next generation of Apple Foundation Models, including a major Siri overhaul expected later this year with iOS 27 and macOS 27. Bloomberg previously reported Apple would pay around $1 billion annually for the arrangement.
That partnership appears to have thawed a relationship that was always a bit stiff. A native YouTube app finally showed up on Apple Vision Pro. And now there's Janus. AppleInsider's coverage frames this as part of a broader détente, with Google becoming a more willing participant in Apple's ecosystem.
There is an irony worth sitting with here. Apple is building its next-generation Siri on top of Google's models, and simultaneously, Google is shipping a native Mac app that will compete directly with that Siri. Both companies seem fine with this arrangement, or at least fine enough to cash the checks.
The "early version" caveat
Google told testers bluntly: "This is an early version of the Gemini for Mac app for your feedback and will have only critical features from the other clients but not all." That's pretty transparent as corporate communications go. The beta is stripped down. No release timeline has been given, and a Google spokesperson wouldn't provide one.
Some speculation (and it is speculation) puts a public launch somewhere between May and September, with Google I/O being the obvious stage for an announcement. But I wouldn't bet on a specific date. Google has a habit of testing things publicly for months before committing to a ship date, and the "critical features only" language suggests this build is far from final.
What's actually in the beta
Testers are working with image and video generation, music creation, tables and charts, math analysis, web search, document uploads, and conversation history. Essentially the same feature set as Gemini on mobile, minus whatever Desktop Intelligence ends up being when it is actually functional. The interface mirrors the iPhone and iPad versions, which is either a sign of design consistency or a sign that the Mac-specific work hasn't started yet. Could be both.
The competitive picture
OpenAI launched its ChatGPT Mac app last year. Anthropic's Claude desktop app has been available even longer, and Cowork has become one of its most-discussed features. Google entering this space now, with a beta that doesn't yet include its headline feature, feels late. Not fatally late, but late enough that the comparison is unflattering.
The counter-argument is distribution. If the Apple partnership proceeds as reported, Gemini technology will be woven into macOS itself through the revamped Siri. A standalone Gemini app then becomes a complementary product rather than the sole point of access. That's a position neither OpenAI nor Anthropic currently holds with Apple's OS.
Whether that matters to users who have already built workflows around ChatGPT or Claude is a different question. And one Google probably should have started answering a year ago.
