Apple stood on stage at WWDC 2026 and unveiled Siri AI. The new assistant promised natural conversation, personal context from emails and photos, onscreen awareness and systemwide actions. Craig Federighi called it “a dramatically more capable and conversational assistant designed to help users find information and get things done throughout the day.” Yet one promised element stayed absent. No mention of user choice among AI models.
That silence didn’t last. Digital Trends reported days later that code in the first iOS 27 developer beta already contains the backend for exactly that feature. Mark Gurman of Bloomberg noted the framework supports models beyond the current ChatGPT integration. Discussions with OpenAI, Anthropic and Google have taken place. The infrastructure sits ready. Apple simply chose not to highlight it.
Why the omission? Several factors likely played a role. Announcing broad third-party access could have muddied regulatory talks in Europe. It risked drawing attention away from Apple’s own foundation models. And it might have invited legal questions around ChatGPT’s current prominence. The story already carried enough complexity. Adding model selection would have required even more explanation.
But the code tells another story. An AI chatbot picker appears in the betas. Right now it toggles between Siri and ChatGPT. That list stands poised to expand. Developers can apply for entitlements. Compatible AI apps could surface in a dedicated App Store section. If the feature ships in the final iOS 27 release, Siri transforms. No longer a single voice. Instead a smart router that directs queries to the model best suited for the task.
Apple’s own models received significant upgrades. The company introduced the third generation of Apple Foundation Models. These include a 3-billion-parameter on-device version and a more powerful 20-billion-parameter multimodal model reserved for newer hardware. CNET detailed how Apple and Google collaborated on these models. The most advanced on-device option delivers expressive voices and sharper dictation accuracy. It runs only on iPhone 17 series devices, M3 or later Macs with sufficient memory, and comparable iPads.
Private Cloud Compute handles heavier lifts. Requests leave the device without storing personal data. Apple says independent experts have verified the system. No logs for Apple or partners to access. This architecture underpins both on-device and server-side processing. It powers the new Siri AI features rolling out in beta to developers now and to users later this year.
The supported devices list remains strict. iPhone 16 models and later qualify, along with iPhone 15 Pro variants. Older iPhones miss out entirely. Macs need M1 chips or better. The same holds for iPads. Apple Vision Pro and recent Apple Watches gain access when paired correctly. Regional limits apply too. European users on iOS will wait longer. China faces regulatory delays.
Yet the multi-model framework could broaden appeal. Users might route complex reasoning to one provider and creative tasks to another. Or stick with Apple’s models for privacy-sensitive work. The choice becomes theirs. And that marks a departure. For years Apple insisted on building every layer itself. Generative AI moves too fast. No single company masters every capability at once.
Recent coverage reinforces the shift. Apple’s press materials emphasize the next-generation foundation models that run locally or through Private Cloud Compute. They highlight conversational depth. Siri AI now pulls relevant details from a user’s messages, suggests calendar entries, even brainstorms recipes based on what’s visible on screen. Visual Intelligence expands across Camera, screenshots and spatial computing on Vision Pro.
Writing Tools appear everywhere. They draft text, refine tone to match a user’s style, proofread automatically. Image Playground generates visuals. Photos gains advanced editing. All of it ties back to the new models. But the beta code hints at something larger. Siri as orchestrator rather than sole performer.
Analysts have watched this evolution. Apple partnered with Google for aspects of the foundation models. A custom Gemini variant reportedly contributes under the hood in some scenarios. Yet the company wraps everything in its privacy layer. The user never sees the handoff. At least not yet.
If the model selector arrives, that changes. A settings screen or quick toggle could let users pick Claude for thoughtful analysis, Gemini for current events or Apple’s own model for on-device speed. Developers would build integrations once. The entitlements process keeps quality in check. An App Store category surfaces the options cleanly.
The timing feels deliberate. WWDC focused on what Apple built. The foundation models. The conversational leap. The personal context that sets Siri AI apart from generic chatbots. Spotlighting third-party access might have undercut that message. So the company held back. The code, however, moved forward.
Betas don’t guarantee shipping features. Apple has cut elements before. But Gurman expressed confidence. “I have little doubt the feature will eventually ship,” he wrote. The backend controls already exist. The UI elements sit in place. Internal builds reportedly enable it fully.
That pragmatism could define Apple’s AI approach going forward. Own the interface. Control the privacy guarantees. Offer choice where it adds value. Competitors race to build the biggest model. Apple bets on making model selection effortless and invisible when desired.
Developers stand to gain too. The Core AI SDK opens new doors. Apps can tap the on-device models at no cost per request. They work offline. Privacy remains central. Combine that with optional third-party routing and the possibilities multiply.
Of course challenges remain. Regulatory scrutiny could intensify if Apple effectively distributes other companies’ AI. Performance differences between models might confuse users. And the most powerful features still require the latest hardware. Not every iPhone owner will see the full picture.
Even so the direction looks clear. Siri AI launches this fall in beta. The foundation sits in iOS 27. And hidden in that update lives the potential for users to pick their intelligence source. Apple didn’t talk about it on stage. The code did the talking instead.
Industry watchers will track the next betas closely. Any expansion of the chatbot picker will signal intent. If the App Store section materializes, the strategy becomes official. Siri stops competing as one model among many. It becomes the front door to all of them.
The bet carries risk. Users might drift to flashier options. Yet it also carries promise. Apple retains the relationship. It controls the experience. And it gives customers what they increasingly demand. Flexibility without sacrificing the privacy and integration that define the platform.
One thing seems certain. The version of Siri arriving later this year will feel different. More capable. More aware. And, if the framework survives, more open to the AI world beyond Apple’s walls.
WebProNews is an iEntry Publication