Meta has unveiled a new open‑source model, and Apple is joining the effort to make it run locally on personal hardware. The two companies say the collaboration will let developers and power users install the model on Macs and iPhones without sending data to the cloud. The announcement was made in a Stratechery post on October 20, 2026.
What the partnership means
Running the model locally means you keep your prompts and outputs on your device. Apple will provide the necessary software hooks and hardware acceleration so the model can run at acceptable speed on its silicon chips. Meta says the model is fully open‑source, so anyone can inspect the code, modify it, or embed it in their own applications.
Limitations and next steps
The post does not give performance numbers, pricing, or a release timeline for the Apple‑enabled version. Meta and Apple have not disclosed which devices will be supported first, nor have they released an independent benchmark. Until more details appear, users will need to wait for the software updates from Apple and the model download from Meta’s repository.
Why it matters
If the partnership delivers, everyday users could run a capable AI model on their phones or laptops without relying on cloud services, preserving privacy and reducing latency. However, until Apple releases the necessary software and Meta publishes the model, it remains unclear how fast or widely the technology will work in practice.
