{"id":4630,"date":"2026-06-06T15:55:46","date_gmt":"2026-06-06T15:55:46","guid":{"rendered":"https:\/\/rentamac.io\/?p=4630"},"modified":"2026-06-06T15:55:54","modified_gmt":"2026-06-06T15:55:54","slug":"ios-27-leaks-report","status":"publish","type":"post","link":"https:\/\/rentamac.io\/pt\/ios-27-leaks-report\/","title":{"rendered":"iOS 27 Leaks Point to a Gemini Powered Siri Before WWDC"},"content":{"rendered":"

With Apple’s WWDC 2026 keynote set for Monday, June 8 at 10 a.m. Pacific, a wave of last-minute iOS 27 leaks says the rebuilt Siri reportedly runs on a custom Google Gemini model, not one of Apple’s own. None of it is confirmed by Apple yet. The claims come from press reports, chiefly a Friday roundup by MacRumors drawing on Bloomberg’s Mark Gurman, plus a separate June 3 report from The Information.<\/p>\n\n\n\n

Here is the tension under the rumor. A company that has sold privacy and on-device AI is reportedly paying about $1 billion a year to license a 1.2-trillion-parameter Gemini model that runs on Nvidia Blackwell B200 GPUs inside Google Cloud. That off-device routing to a rival’s cloud is the contradiction Monday’s keynote will have to square.<\/p>\n\n\n\n

Siri’s new brain reportedly belongs to Google<\/h2>\n\n\n\n

The new Siri reportedly runs on a custom Gemini model roughly eight times larger than Apple’s own cloud model, according to The Information by way of Tech Times. Demanding queries route to Google Cloud, while simple requests like timers and smart-home commands stay on the device.<\/p>\n\n\n\n

<\/th>Apple’s own cloud model<\/th>Licensed Gemini model<\/th><\/tr><\/thead>
Parameters<\/td>~150 billion<\/td>~1.2 trillion<\/td><\/tr>
Where it runs<\/td>Apple Private Cloud Compute<\/td>Google Cloud<\/td><\/tr>
Hardware<\/td>Apple silicon<\/td>Nvidia Blackwell B200<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

*Reported figures, not Apple-confirmed.*<\/p>\n\n\n\n

Apple reportedly pays around $1 billion a year for the license. Analyst Gene Munster of Deepwater pegged the full deal nearer $5 billion, an estimate, per Tech Times. TechRepublic frames the arrangement as a break from Apple controlling its full stack, and reports the company abandoned running Gemini-class models on its own Private Cloud Compute over performance limits.<\/p>\n\n\n\n

How Apple is trying to square the privacy gap<\/h2>\n\n\n\n
\"Routing<\/figure>\n\n\n\n

The B200 uses confidential computing that “encrypts all data held in GPU memory, including model weights, the user’s input, and the inference result,” according to The Information via Tech Times. That is the technical argument outlets expect Apple to lean on.<\/p>\n\n\n\n

The open question is whether it lands. Apple’s on-device-AI marketing sits awkwardly next to sending your hardest queries off to Google, and the reported decision to drop its own Private Cloud Compute attempt reads as an admission of a gap in the stack.<\/p>\n\n\n\n

The new Siri may launch behind a waitlist<\/h2>\n\n\n\n

The revamped assistant may ship behind a waitlist, where you sign up and wait for access, a setup multiple outlets agree on. Apple has done this before, running an Apple Intelligence waitlist back in iOS 18.1.<\/p>\n\n\n\n

Apple has also labeled the new Siri “beta” and “preview” internally, suggesting “the assistant won’t be marketed as fully finished software when it’s released later this year,” per Bloomberg via 9to5Mac. How long the waitlist lasts, and which features it gates, is unstated.<\/p>\n\n\n\n

What the leaks mean for Macs and older iPhones<\/h2>\n\n\n\n

The same leak cycle carries a few concrete changes worth tracking:<\/p>\n\n\n\n