Apple Siri AI Public Beta Runs on Google Gemini Models

Apple shipped Siri AI in public beta on September 14, 2026, alongside iOS 27. The assistant runs on foundation models Apple built in collaboration with Google's Gemini team — the first time a non-Apple model has powered a shipping version of Siri (Apple Newsroom).

The beta is open to anyone with a compatible device. English is the only language at launch; French, Japanese, Korean, Portuguese, and Spanish follow in October.

Which Devices Run Siri AI

Siri AI requires an iPhone 15 Pro or newer, an iPad with an M1 chip or later, a Mac with an M1 or later, an Apple Watch Series 9 or later, or Apple Vision Pro. It ships across iOS 27, iPadOS 27, macOS 27, watchOS 27, and visionOS 27. The previous Siri remains available as a fallback on unsupported hardware (Apple Newsroom, Pondero).

Apple's most advanced on-device model, AFM Core Advanced, powers expressive voices and systemwide dictation on a narrower set of devices: iPhone 17 Pro and later, iPad with M4 or later and at least 12GB of unified memory, Mac with M3 or later and at least 12GB, and Apple Vision Pro with M5. Older supported devices still run Siri AI, but without the enhanced voice and dictation features (Apple Newsroom).

The Google Gemini Partnership

Google confirmed the collaboration in April 2026, describing itself as Apple's "preferred cloud provider" for developing next-generation Apple Foundation Models. Apple created smaller derivatives of Google's Gemini models through a distillation process, then tuned them for Apple Silicon. The result gives Siri access to a substantially larger model baseline than Apple's own on-device work had previously delivered (Pondero).

Siri AI uses a system orchestrator that can access the Spotlight index and App Toolbox — both of which run entirely on device. Processing splits between on-device computation and Apple's Private Cloud Compute servers. When PCC handles a request, Apple states that the user's personal data "is not stored, and is not made accessible to Apple or anyone else." Outside experts can verify this claim at any time, Apple says (Apple Newsroom).

That privacy boundary matters because Google-derived models sit in the processing pipeline. Enterprise teams deploying Apple devices in regulated environments may need to assess whether Apple's infrastructure guarantees satisfy their threat models before enabling Siri AI for sensitive workflows (Pondero).

Cross-App Context and On-Screen Awareness

Siri AI can search across Messages, Mail, and Photos simultaneously, read on-screen content, draft emails, edit and share photos, and create calendar events from natural-language requests.

Apple demonstrated the cross-app capability with a family-visit example: a user asks what a relative suggested doing during an upcoming visit. Siri searches Messages, finds that the relative mentioned baking a family recipe, retrieves the cooking instructions from a previously sent email, and adds the ingredients to the user's grocery list in Reminders — all without the user switching apps (Apple Newsroom).

On-screen awareness works similarly. A user viewing a sports team's website can ask Siri to find all upcoming home games, and Siri adds them directly to Calendar. On Mac, users can Control-click images, files, or text and ask Siri about them. On Apple Vision Pro, users can look at content — inside app windows or at physical objects around them — and ask Siri questions (Apple Newsroom).

Siri AI also supports systemwide actions inside third-party apps. Users can text a friend in WhatsApp, play a book on Audible, draft an email in Microsoft Outlook, search for a homework assignment in Notability, or add a restaurant recommendation to Tripsy — all through voice or text requests to Siri (Apple Newsroom, AI Weekly).

That cross-app access is something standalone AI assistant apps on iOS have historically struggled to achieve. iOS sandboxing forced third-party assistants to work around system restrictions. Siri AI, built into the OS, bypasses those limitations entirely (Pondero).

Visual Intelligence and Multimodal Capabilities

Siri AI includes Visual Intelligence, giving the assistant image understanding and multimodal capabilities across iPhone, iPad, Mac, and Apple Vision Pro. On iPhone, a new Siri mode in the Camera app lets users point the camera at something — a music festival poster, for example — and ask Siri for set times. Siri then creates calendar events automatically. On iPad, Visual Intelligence integrates into the screenshot experience. On Mac, users access it through a dedicated keyboard shortcut (Apple Newsroom).

The Siri App and Conversation History

A new dedicated Siri app lets users revisit past conversations and start new ones. Conversation history syncs across devices via iCloud, so a user can start chatting with Siri on iPhone and continue on Mac, iPad, Apple Watch, or Apple Vision Pro. Live Rewind text snippets and Siri Recap summaries are end-to-end encrypted in the Siri app — Apple states these are not accessible even by Apple itself (Apple Newsroom).

Siri AI is not available for users under 13.

Audio Intelligence on Apple Watch

Later in 2026, Apple Watch Series 12 and Apple Watch Ultra 4 will gain Audio Intelligence features, including Live Rewind, Siri Recap, Sound Recognition, and faster Shazam. Live Rewind captures the previous 15 seconds of a conversation as a text snippet via a double-press of the Digital Crown. Siri Recap generates high-level summaries of conversations with titles and key points in the Siri app.

Audio Intelligence does not create or store audio recordings. The S11 chip includes a Secure Exclave — a hardware-isolated compartment that processes audio in complete isolation from the rest of the system. Raw audio is deleted immediately after processing. Audio Intelligence does not identify or attribute speakers. Live Rewind and Siri Recap arrive in beta in late 2026, in English only, and will not initially be available in the EU (Apple Newsroom).

Writing Tools and Image Playground

Siri AI includes integrated Writing Tools that work virtually anywhere text input is available. Users can describe what they need and Siri generates a draft, or describe a change and Siri updates existing text. In Mail and Messages, Siri can mirror the user's typical communication style — including punctuation and tone — for each recipient. Apple Intelligence also proofreads automatically as users type, including in most third-party apps (Apple Newsroom).

Image Playground, which runs on Private Cloud Compute, gains photorealistic image generation and extends into the Siri app. Users can generate images by describing them in the Siri app. SynthID watermarking support will arrive in a software update later in 2026 to identify images generated or edited with Apple Intelligence tools (Apple Newsroom).

Where Siri AI Is Not Available

The European Union is excluded from the iOS, iPadOS, and watchOS launch. Apple says it is "working to find a path forward that preserves user privacy and security." China remains on hold pending regulatory approval. The company has not provided a timeline for either region (Apple Newsroom, AI Weekly).

Daily Usage Limits and Future Pricing

Apple Intelligence features that rely on server-side models — including Siri AI, Image Playground, intelligent photo editing tools, AFM 3 Cloud models in Shortcuts, Live Rewind, and Siri Recap — are subject to daily usage limits. Limits may vary based on the feature, request complexity, system demand, and system policies. Apple states that "expanded access to these features will be available for a fee in the future," though no pricing has been announced (Apple Newsroom).

How to Join the Siri AI Beta

Update to iOS 27, iPadOS 27, macOS 27, watchOS 27, or visionOS 27.
Enroll in Apple's Beta Software Program if you haven't already.
Open Settings > Siri & Search.
Toggle Siri AI on.
Try a cross-app request like "What did Sarah send me about the dinner plans?"

What's Next

Apple has not announced a general-availability date for Siri AI. The beta runs in parallel with fall hardware refreshes expected through September and October. Five additional languages arrive next month. The EU regulatory outcome could unlock tens of millions of additional devices — or keep Siri AI off Apple's second-largest market indefinitely. And Apple's statement that expanded server-side access will eventually require a fee raises a question the company has yet to answer: how much will full Siri AI cost, and will the free tier be enough for daily use?

Sources: Apple Newsroom, AI Weekly, Pondero, AI Weekly Daily Edition