iOS 27's Rebuilt Siri AI Handles 125 Real-World Tasks in Independent Test

Independent Testing Shows Broad Capability Gains in iOS 27's Siri AI

Apple's rebuilt Siri AI, shipped this week in iOS 27, passed nearly all of 125 real-world tasks in an independent test conducted by journalist David Pogue, according to a report from 9to5Mac. Pogue published his findings in a Substack post titled "125 Tests of the New AI Siri," documenting prompts ranging from simple reminders to complex, context-dependent queries.

What Was Tested

Pogue's list covers commands well beyond Siri's historical scope of timers, reminders, and music playback. Examples cited by 9to5Mac include:

• Cross-app task aggregation: Asking "What are people waiting on me for?" prompted Siri to generate a consolidated list pulled from Mail, Messages, and Reminders — surfacing an item Pogue said he had genuinely forgotten about.
• Contextual awareness without explicit details: Asking "What's the apartment number!?" while running late to a party returned the correct address ('The apartment number for the Bynum party tonight is 10-F') despite Pogue never specifying which party or date in his prompt.
• Bulk actions with confirmation: A request to "Delete all my alarms" triggered Siri to first confirm to delete 38 alarms in the Clock app before executing.

According to 9to5Mac, Siri AI succeeded at nearly every one of the 125 tests. The outlet reports that most failures — such as sending a Venmo payment — stemmed from third-party apps not yet updated for iOS 27 and Siri AI compatibility, rather than limitations in Siri itself.

Confirmed Facts vs. Claims

The 125-test count, specific prompt examples, and success rate are drawn from Pogue's published Substack post as relayed by 9to5Mac; Apple has not released its own benchmark or accuracy figures for Siri AI's task completion rate. No technical specifications — such as underlying model architecture, parameter count, or whether Siri AI runs on-device, in Apple's private cloud, or via a third-party model partnership — were disclosed in the source reporting.

9to5Mac's Ryan Christoffel noted that longtime iPhone users have historically limited their Siri usage to a narrow set of reliable tasks, such as reminders and messaging while driving, due to years of inconsistent performance on more complex requests.

What This Means

This is a product capability report, not a model release with disclosed weights, benchmarks, or pricing. The significance lies in the qualitative leap Pogue describes: Siri moving from single-intent commands to multi-app reasoning and implicit context resolution — capabilities that put it in closer competition with assistants like Google's Gemini and OpenAI's ChatGPT voice mode. The real test will be third-party app compatibility, since several of Pogue's noted failures depend on developers updating their apps for iOS 27's new Siri integration rather than on Siri's own reasoning ability. Apple has not published independent benchmarks, so claims of near-universal success rest on one journalist's manual testing rather than a standardized evaluation.