Home · Technology · Oct 15 archive
Meta and Apple Collaborate on Muse Spark OpenChatGPT for Macs
Confirmed
In Short: Meta and Apple are working together to optimize Meta's new Muse Spark Open AI model for high-end Macs, offering developers a powerful tool for coding and complex tasks.
Meta and Apple have announced a collaboration to optimize Meta's new Muse Spark Open AI model for high-end Macs, allowing developers to run the model locally without relying on cloud servers.
The Muse Spark Open model, which boasts 816 billion total parameters, is designed to excel in long-horizon tasks, complex software engineering, and multi-step reasoning.
According to Meta, the model was optimized to run on Mac Studio with M5 Ultra, and two Mac Studios can behave like one supercomputer with a terabyte of unified memory.
Apple's MLX framework splits the model across connected Macs, which communicate over Thunderbolt 5, enabling efficient processing.
Developers need either two Mac Studios with 512GB of unified memory each or four with 256GB each to run the model, with a total cost exceeding $40,000.
In Apple's tests, four Mac Studios generated more than 36 tokens per second, showcasing the model's performance.
The Muse Spark Open model's weights are available for download on Hugging Face, Amazon SageMaker AI, and Google Cloud Model Garden.
Meta CEO Mark Zuckerberg announced the collaboration on X, expressing excitement about the model's performance on Mac Studios.
Apple CEO John Ternus praised the collaboration, noting that it pushes the boundaries of on-device AI.
The companies have been collaborating for several months, and representatives declined to comment on whether this is the start of a deeper relationship.
What this adds
The hardware requirements for running the Muse Spark Open model differ between sources, with some accounts suggesting the need for more Mac Studios.
Meta's Muse Spark Open model is described as the most capable US open-weight model to date, offering transparency and flexibility that closed-source models do not.
The collaboration between Meta and Apple marks a significant step in the development of on-device AI capabilities, potentially reducing reliance on cloud services for complex AI tasks.
Background
Meta has launched the Muse Wonder AI model, a follow-up to its earlier Muse Spark models, as part of its ongoing efforts to revitalize its artificial intelligence division.
Meta has launched its MuseOpenAI's ChatGPT on the iPad, expanding its availability beyond the iPhone and Mac.
What's confirmed
- “Built for coding and agentic work, it excels at long-horizon tasks, complex software engineering, and multi-step reasoning,” says Meta.
What's still developing
- By Ed Hardy. 2:35 pm, October 15, 2026 Meta and Apple are working together to bring a powerful new open-weight AI model to high-end Macs.
- An open-weight AI model is a type of artificial intelligence whose core “building blocks” — the internal settings and patterns learned during training — are made freely available for anyone to use.
- By making these internal components accessible, open-weight models offer much greater transparency, allowing users to better understand how decisions are made by the AI.
- “Leveraging the latest version of MLX, Muse Spark Open was optimized to run on Mac Studio with M5 Ultra, and together they’re the best way to run a powerful model on your own desktop,” said Meta in a statement.
- Apple 's 2021 privacy changes curtailed ad targeting, and Meta put the resulting revenue hit at $10 billion for 2022, a figure it disclosed in early 2022, according to CNBC.
- "They should be American." Accounts of the hardware needed differ.
- "Most of the world's AI will run on open models," Wang told Bloomberg.
- REGISTER NOW Meta has launched what it calls the “most capable” open-weight model in the United States.
- The new model, which leverages 816 billion parameters and a 1 million token context window, claims to offer superior performance to other popular open-weight models, according to evaluation results released by Meta on Thursday.
- Open-weight models provide a number of distinct industry advantages that aren’t afforded by closed and proprietary models.
- Still, if you can muster the memory, Muse Spark Open may prove a significant work horse for coding and other desktop tasks.
- Open-weight models can be significantly more cost-efficient and can also offer transparency and flexibility that closed source models don’t allow.