Apple just dropped an update on its AI game—and let’s just say, the reviews are… mixed.
In a recent blog post, Apple introduced its latest AI models—Apple On-Device and Apple Server—claiming smoother performance, smarter tools, and wider language support. But dig a little deeper, and things don’t look quite as polished.
According to Apple’s own benchmark tests, these shiny new models still fall behind the competition. The On-Device model? Human testers said it’s on par with similar models from Google and Alibaba—but not better. And Apple Server, the heavyweight model designed to run from data centers? It landed behind OpenAI’s GPT-4o… which has been around for over a year.
Ouch.
It doesn’t stop there. Apple also admitted that in visual analysis tasks, its model was beaten by Meta’s Llama 4 Scout—even though that model ranks lower on several fronts when compared to Google, OpenAI, and Anthropic models.
Let’s be real—this adds fuel to the growing perception that Apple is trailing in the AI arms race. From delayed Siri upgrades to underwhelming user experiences, the world’s most valuable company seems to be playing catch-up while the rest of the AI field is sprinting ahead.
Still, Apple is pushing forward. The On-Device model (clocking in at 3 billion parameters) powers features like text summarization and offline AI tools on iPhones and Macs. It’s compact, works offline, and is now accessible to third-party developers through Apple’s Foundation Models framework.
Both models also benefit from expanded training datasets, including image data, tables, and PDFs—suggesting Apple’s trying to build models that are more versatile and multi-modal.
But here’s the bigger question: Is this enough?
While Apple touts improved efficiency and multilingual capability (supporting about 15 languages), the tech world is rapidly evolving—and “almost good enough” doesn’t win market share anymore.
In an era where startups are building apps with 10-person teams and AI agents are doing the work of full engineering departments, Apple needs to deliver more than brand loyalty and hardware polish. If it wants to compete in this space, it’s going to need more than just sleek marketing. It needs results.
Because in the AI world, speed—and performance—is everything.
