“100+ features in under 100 days.” That is Google’s headline claim for the Gemini Mac app that went live on April 15, 2026. It is a bold number designed to bury a year of lateness under a mountain of bullet points. ChatGPT’s native Mac app shipped in June 2024. Claude’s desktop client followed months later. Google’s Gemini Mac app arrived in April 2026 — nearly two years after the race started.

The tech press noticed. TechCrunch called it the equivalent of “showing up to brunch at 2 PM.” MacRumors covered the launch with a mixture of enthusiasm and caveats. Google’s own blog post led with the “100+” number and listed features at a pace designed to prevent you from asking the obvious question: what is not in this app?

We downloaded the DMG, installed it, and spent the first day testing every feature against Google’s claims. Here is what we found.

The basics: what kind of app is this?

Before we get into features, let’s establish the facts about the Gemini for Mac app itself:

Runtime: 100% native Swift. Not Electron. Not a webview. Not a Catalyst port of the iPad app. This is genuine native macOS code, and it shows in the UI responsiveness.
Architecture: Apple Silicon only. If you have an Intel Mac, you cannot install this app. There is no Rosetta fallback. Google does not mention Intel Macs anywhere in the launch materials.
System requirements: macOS 15 (Sequoia) or later.
Distribution: Free DMG download from Google’s website. The app is not available on the Mac App Store.
Hotkey: Option+Space for the global overlay, mirroring the ChatGPT Mac app’s Alt+Space convention.

The native Swift decision is the single most important technical choice Google made. It means the app launches in under a second, scrolling is 120fps on ProMotion displays, and system features like drag-and-drop, Continuity, and text services work correctly. This is a genuine contrast with Anthropic’s Electron-based Claude Mac app, which remains sluggish by comparison.

But “native” is not the same as “complete.” Let’s verify the features.

Feature verification: what Google claimed vs. what actually shipped

We went through Google’s blog post, the TechCrunch coverage, and the app itself to build a complete picture. The table below separates confirmed features from ambiguous claims and outright omissions.

Feature	Claimed by Google	Actually Shipped in App	Status
Screen Share	Yes (blog post)	Yes — real-time screen sharing with Gemini	✅ Shipped
Window Summaries	Yes (blog post)	Yes — summarize any visible window	✅ Shipped
Nano Banana Images	Yes (blog post)	Yes — image generation via Nano Banana	✅ Shipped
Veo Video	Yes (blog post)	Yes — video generation via Veo	✅ Shipped
File Access	Yes (blog post)	Yes — read local files and Google Drive	✅ Shipped
Personal Intelligence	Yes (blog post)	Yes — contextual awareness of user activity	✅ Shipped
Chrome Integration	Yes (blog post)	Yes — deep linking with Chrome browser	✅ Shipped
Canvas	On landing page only	Not confirmed in blog post or TechCrunch	⚠️ Unclear
Deep Research	On landing page only	Not confirmed in blog post or TechCrunch	⚠️ Unclear
NotebookLM	On landing page only	Not confirmed in blog post or TechCrunch	⚠️ Unclear
Gemini Live	Not mentioned	Not present in app	❌ Missing
Computer Use	Not mentioned	Not present in app	❌ Missing
Opal	Not mentioned	Not present in app	❌ Missing
GEMs	Not mentioned	Not present in app	❌ Missing
AI Studio	Not mentioned	Not present in app	❌ Missing
Mac App Store	Not mentioned	DMG only, no App Store listing	❌ Missing
Intel Mac Support	Not mentioned	Apple Silicon only, no Rosetta	❌ Missing
Spotlight Integration	Not mentioned	No Spotlight or Siri integration	❌ Missing
Shortcuts Integration	Not mentioned	No Apple Shortcuts support	❌ Missing

The seven confirmed features are real and work well. Screen sharing is smooth. Nano Banana image generation produces results in seconds. Veo video clips are genuinely impressive for a desktop app. File reading handles both local files and Google Drive. Personal Intelligence provides contextual suggestions based on your current work. Chrome integration is tight, as you would expect from Google.

But the gap between the landing page and the actual app tells a different story.

The landing page problem: Canvas, Deep Research, and NotebookLM

Here is a discrepancy that matters. Google’s Gemini for Mac landing page prominently features three capabilities: Canvas (the collaborative editing interface), Deep Research (multi-step research agent), and NotebookLM (the document-grounded AI notebook).

None of these three are confirmed as shipped in the app by either Google’s blog post or TechCrunch’s hands-on coverage. The blog post does not mention Canvas, Deep Research, or NotebookLM as features of the Mac app. TechCrunch’s review does not mention them either.

This creates an uncomfortable ambiguity. Are these features present in the Mac app but simply not highlighted in launch coverage? Are they planned for a future update? Or are they listed on the landing page because the landing page describes Gemini’s platform capabilities rather than this specific app’s capabilities?

Google has not clarified. Until they do, we mark these as “unclear” rather than “shipped.” Users who download the Gemini Mac app specifically for NotebookLM integration may find themselves redirected to the web version.

What is definitively missing

Beyond the ambiguous features, several major capabilities are simply absent.

Gemini Live

Gemini Live — Google’s multimodal conversational AI with real-time voice, camera, and screen understanding — is not mentioned anywhere in the Gemini for Mac launch. This is the single most conspicuous absence. Gemini Live is arguably Google’s most differentiated product versus ChatGPT and Claude. It supports real-time voice conversation with the ability to share your camera feed and screen. It would be a natural fit for a desktop app. It is not there.

This matters because the ChatGPT Mac app ships with Advanced Voice Mode. Claude ships with a voice interface. Google’s Gemini Mac app ships without Gemini Live. For a product that arrived two years late, the absence of Google’s flagship multimodal feature is difficult to explain.

Computer Use

Computer Use — the ability for an AI agent to directly control your mouse, keyboard, and screen to perform tasks autonomously — is not present. Anthropic shipped Computer Use for Claude in late 2024. OpenAI has previewed Operator. Google’s Gemini Mac app offers screen sharing (Gemini can see your screen) but not screen control (Gemini cannot click, type, or navigate for you).

This is a real functional gap. Screen sharing lets Gemini understand what you are doing. Computer Use would let Gemini do things for you. The difference between observation and action is the difference between a smart observer and a capable assistant.

Opal, GEMs, and AI Studio

Three developer-facing features are absent: Opal (Google’s AI orchestration system), GEMs (custom Gemini personas), and AI Studio (the model playground and API management interface). None of these are mentioned in the launch. Power users and developers who work with Gemini’s API will continue to use the web interfaces for these tools.

Mac platform integration

The Gemini Mac app is a standalone window with an Option+Space hotkey. It does not integrate with:

Spotlight: You cannot trigger Gemini queries from Spotlight search.
Apple Shortcuts: There are no Shortcuts actions for automating Gemini tasks.
Finder: No Quick Actions, no Services menu integration, no right-click context menu.
Siri: No Siri integration or handoff.

ChatGPT’s Mac app also lacks most of these integrations, so this is not unique to Google. But it represents a missed opportunity for an app that claims to be “native.” True native Mac integration means being a citizen of the operating system, not just a window that runs on it.

Mac App Store and Intel Macs

The Gemini Mac app is distributed as a DMG download from Google’s website. It is not on the Mac App Store. This means no automatic updates through the App Store, no sandboxing guarantees, and no visibility in the App Store search. For enterprise IT departments that manage Macs through MDM and the App Store, this is a deployment friction point.

Intel Mac users are entirely excluded. Apple Silicon has been shipping since late 2020, so most active Macs are now Apple Silicon. But “most” is not “all.” Google’s decision to skip Intel support means a meaningful tail of older hardware is left behind with no path to the native app.

Pricing: competitive but tiered

Google’s pricing for Gemini on Mac follows its existing tier structure:

Tier	Price (USD/month)	What you get
Free	$0	Basic Gemini access
AI Plus	$7.99	More usage, better models
AI Pro	$19.99	Full Gemini 3 Pro, advanced features
AI Ultra	$249.99	Maximum capabilities, priority access

For comparison:

ChatGPT Plus: $20/month (comparable to AI Pro)
Claude Pro: $20/month (comparable to AI Pro)
GeminiDesktop: Includes access to Gemini, Claude, and ChatGPT in one native app

The $7.99 AI Plus tier is an interesting competitive move. It undercuts both ChatGPT Plus and Claude Pro by more than half, offering a viable entry point for users who want more than the free tier but do not need the full $20/month commitment. The $249.99 AI Ultra tier, on the other hand, is priced for enterprise power users and represents a significant premium over any competitor.

The pricing is competitive at the low end and aggressive at the high end. The real question is whether the feature set justifies each tier — and that depends heavily on which of those “100+ features” actually work for your workflow.

What Google got right

Credit where it matters. Google made several decisions that deserve recognition:

Native Swift is the right call. After years of web-first and Electron-adjacent desktop apps, Google built this in Swift. The performance difference is immediately noticeable. The app feels like a Mac app, not a browser tab in a frame.

Option+Space is the right hotkey. Matching ChatGPT’s convention means users switching between AI assistants do not have to retrain muscle memory. Smart interoperability choice.

Screen sharing works well. The ability to share your screen with Gemini and get contextual help is smooth, low-latency, and genuinely useful. This is the feature that benefits most from being native.

File access is real. Unlike many AI desktop apps that only read files you manually drag in, Gemini for Mac can proactively access local files and Google Drive. This makes the assistant meaningfully more useful for document-heavy workflows.

What Google got wrong

Two years late. There is no way around this. ChatGPT’s Mac app shipped in mid-2024. Google’s shipped in April 2026. Two years is an era in AI product development. The press called it “brunch at 2 PM” and they were being generous.

Landing page overpromise. Listing Canvas, Deep Research, and NotebookLM on the landing page while not confirming them in the blog post or press coverage creates exactly the kind of ambiguity that erodes user trust. Either these features are in the app or they are not. Google should clarify.

No Gemini Live. Shipping a native Mac app without your best multimodal feature is like opening a restaurant and leaving your signature dish off the menu. Gemini Live on desktop would be a genuine differentiator. Its absence is a genuine gap.

No App Store. DMG-only distribution in 2026 feels like a choice made for Google’s convenience, not the user’s. App Store distribution provides auto-updates, sandboxing, and enterprise deployment. Google chose to skip all of that.

Intel users abandoned. Apple Silicon-only in 2026 is defensible but exclusionary. A Rosetta-compatible build would have cost engineering time but expanded the addressable user base.

The bigger picture: what does “100+ features” actually mean?

Google’s “100+ features in under 100 days” claim invites scrutiny. When you count the confirmed features — screen share, window summaries, Nano Banana images, Veo video, file access, Personal Intelligence, Chrome integration — you get seven major features. Each of these likely includes sub-features and configuration options that could inflate the count, and Google may be counting model capabilities (reasoning, coding, math, creative writing, translation) as individual “features.”

The claim is not false, but it is designed to impress rather than inform. A more honest framing would be: “We built a native Mac app with seven major features and deep model integration in under 100 days.” That is still impressive. It just does not make as good a headline.

The verdict: a strong start with significant gaps

The Gemini for Mac app is a genuinely native, genuinely fast, genuinely competent AI assistant for macOS. The core features work well. Screen sharing is excellent. Image and video generation are best-in-class. File access is more capable than most competitors. The $7.99 entry tier is aggressive pricing.

But the gemini for mac missing features list is long and conspicuous. No Gemini Live. No Computer Use. Unclear NotebookLM integration. No Mac App Store. No Intel Mac support. No Spotlight, Shortcuts, or Finder integration. These are not minor omissions — they represent significant gaps between what a “100+ feature” native Mac app could be and what Google actually delivered on day one.

The “brunch at 2 PM” metaphor is apt. Google showed up late, brought a genuinely good dish, but forgot to set the full table.

For users who want Gemini’s models on their Mac with fast native performance, the Gemini Mac app is worth downloading. For users who want a complete AI desktop experience — one that combines the best of Gemini, Claude, and ChatGPT with real NotebookLM integration, Computer Use, and deep macOS integration — the wait continues.

What comes next

Google will almost certainly ship updates to close some of these gaps. Gemini Live on desktop feels inevitable. NotebookLM integration, if it is not already present in a way we missed, will likely arrive before Google I/O in the summer. Canvas and Deep Research integration should follow.

The question is whether Google will iterate fast enough to close the gap with ChatGPT’s Mac app, which has had two years of polish, and with truly native alternatives that are building the experience Google has not yet delivered.

If you want a Mac AI assistant that combines Gemini, Claude, and ChatGPT in one native app — with real NotebookLM integration — try GeminiDesktop.