Craig Federighi, Apple’s SVP of Software Engineering, held a post-keynote press briefing on Monday (reported by 9to5Mac) alongside AI VP Amar Subramanya, Siri lead Mike Rockwell, and software VP Sebastien Marineau-Mes. The group walked through the development of the third-generation AFM family and how it powers Apple Intelligence.
“The amount of Google Assistant we use is none,” Federighi said, emphasizing that Apple relies on none of the Gemini models Google deploys to its customers, none of Google’s client-side code, and no Google Search infrastructure as its knowledge backbone.
Of course, we don’t have the Gemini app as our app. In fact, none of that client code is part of how we run on iOS. For these models, we use none of the models that Google deploys to their customers, nor do we use the infrastructure and means by which they deploy models to their customers. And then, when it comes to the knowledge base, we of course don’t use Google Search or anything like that as the foundation of our system. — Craig Federighi, Apple’s SVP of Software Engineering
Hartley Charlton for MacRumors:
Subramanya outlined the new AFM family, which spans two on-device models and three server-side models. The on-device tier consists of AFM Core, a next-generation dense architecture model, and AFM Core Advanced, which uses a sparse architecture and is natively multimodal.
Subramanya said AFM Core Advanced is “unlike any on-device model we’ve run before,” enabling new features including invitation and expressive voices without any cloud requests. On the server side, AFM Cloud handles latency-optimized Private Cloud Compute requests, while AFM Cloud Image powers image generation and editing features including spatial reframing.
The key detail on the Google collaboration came in Subramanya’s description of how these four models were trained. “All of these are custom built for Apple Silicon, trained using proprietary data with reinforcement learning and refined using outputs from Gemini frontier models,” he said, making clear that Google’s contribution was distillation-based, not a wholesale adoption of Gemini.
The fifth and most capable model, AFM Cloud Pro, is designed for agentic tool use and complex reasoning tasks, with quality that Subramanya said is “similar to Gemini frontier models.” This model marks a departure from Apple’s standard Private Cloud Compute setup.
To run it, Apple worked with both Google and Nvidia to extend its private cloud infrastructure to Nvidia GPUs hosted in Google’s cloud. Marineau-Mes said Apple wanted to use Nvidia’s latest chips but required them to be configured so they couldn’t read the contents of Apple’s servers. A recent Nvidia technology called “ambiguous confidential compute” provided the solution.
MacDailyNews Note: Read more in the full article here and also see our own article posted yesterday: Apple’s Siri AI is not ‘Google Gemini with Apple branding’; here’s how it really works.
Please help support MacDailyNews — and enjoy subscriber-only articles, comments, chat, and more — by subscribing to our Substack: macdailynews.substack.com. Thank you!
Support MacDailyNews at no extra cost to you by using this link to shop at Amazon.
