Search

The good news is that you don't need a massive, GPU-melting frontier model for this workflow. Smaller models are perfectly capable of identifying missing context and asking useful follow-up questions. Popular options include lightweight open models like Google's Gemma 4, Meta's Llama 4 Scout, Microsoft's Phi-4, and compact models from Mistral and Qwen. These models are readily available as mentioned through tools like LM Studio and GPT4All, running comfortably on standard consumer hardware.

The list of devices that support all of the AI features is fairly limited. Here's the full range of supported devices: iPhone 17 ProiPhone 17 Pro MaxiPhone AiriPad models with M4 and later and at least 12GB of unified memoryMac models with M3 and later and at least 12GB of unified memoryApple Vision Pro (M5) That means even if you have the popular iPhone 16 Pro or Pro Max and all the Apple Intelligence features it supports, you'll be missing out on some of the new features. Sure, these devices support iOS 27, but some things will be left out.

After testing both approaches, I don't think cloud AI and local AI are actually competitors. Using local AI has taught me interesting tricks with cloud AI and vice versa. They're really just different tools for different situations. If you want the smartest models, built-in web access and the easiest user experience, cloud AI is still the clear winner. If privacy, customization and offline access matter most, local AI offers advantages that cloud platforms simply can't match. Personally, I've found myself using both. I rely on cloud AI for research, brainstorming and complex projects. But for

Apple didn't provide a full rundown of the features that will be left off older devices, but, as noted in Craig Federighi's quote, "expressive voices and more advanced dictation" are part of the on-device functionality limited to new devices. The company demonstrated the expressive voices feature during its keynote; essentially, it lets you fine-tune how Siri sounds. You can adjust the pitch and speed at which the personal assistant speaks. While it sounds useful, it's hardly a game-changer. The more advanced dictation wasn't dove into as deeply, but Apple did say that it "now captures what us