It’s the enemies-to-lovers trope, tech edition. Two tech titans have teamed up to create a new open-weights AI model. Meta is the owner of the new AI model, called Muse Spark Open, and it worked with Apple’s hardware team to ensure the AI model can run efficiently on the M5 Mac Studio Ultra.
Muse Spark Open is an open-weights AI model, which means developers can understand how it works and customize it to their needs. It has a robust 42 billion active parameters and a context window of a million tokens, with multimodal capabilities that Meta says are good at agentic and coding tasks. Translation: It’s a big model that can handle different kinds of tasks, including powering AI agents, aimed at engineers and developers. The model’s weights are publicly available now on HuggingFace under Meta’s Muse Model license.
Note that if you want to run the model on Apple computers, it requires two M5 Mac Studio Ultras to run, so at $5,499 each, it’s a pricey investment.
More from CNET
Meta and Apple could not have taken more divergent paths in the AI race, or even in the pre-AI era. Apple has been hesitant to throw itself into AI development, whereas Meta CEO Mark Zuckerberg couldn’t push his company fast enough, burning through billions of dollars to do so. Apple’s decades-long focus on privacy, while not perfect, is existentially at odds with Meta’s entire targeted advertising business model reliant on collecting as much information on its users as possible. Remember when Apple offered iPhone owners the ability to opt out of app tracking and cost Meta $10 billion in 2022?
But this new collaboration is a clear sign that both companies are determined to travel down their set paths in the AI race: Meta, to build open-weights AI models, and Apple, to build the hardware needed to run AI. Or as Alexandr Wang, Meta’s chief AI officer, told Axios, “We’re excited for mommy and daddy to be playing nice again.” In X posts, Zuckerberg put it a bit more diplomatically: New Apple CEO John Ternus focused on the reason for the team-up, namely on-device AI.
On-device AI is a way to run AI models and tools without relying on a cloud service. It requires a little more setup than simply opening up your ChatGPT app, but developers and businesses can configure the AI system to keep their data more secure. Depending on the scale of the operation and hardware used, it can also be a cheaper, long-term option than continually paying a monthly subscription.
It’s not totally surprising to see Apple taking part in AI development, despite its own meager AI tools. The Cupertino company has carved out a place in the AI race through its hardware. At the beginning of 2026, when tech circles were excitedly using the OpenClaw agent, Apple’s Mac Minis were the computer of choice for many AI fans. During Apple’s hardware launches, including the M5 MacBook Air in March, Apple boasted that its chip tech could power impressive performance, and CNET’s testing mostly verifies those claims. The Mac Studio, released in August, is the latest example of Apple focusing on hardware capable of running compute-intensive AI tasks.