Muse Spark Open runs locally on Apple silicon, marking a rare direct collaboration between two companies with competing AI agendas.
Meta released Muse Spark Open, an open-weight model co-engineered with Apple to run natively on Mac hardware. Meta is calling it the most capable US open-weight model released to date, and the first in its class optimized for Apple's MLX framework rather than cloud inference.
The partnership is notable because Apple has largely kept its AI stack closed and has been slow to ship competitive in-house models. Pairing with Meta's open-weight strategy gives Apple a credible on-device AI story without building a frontier model itself.
For operators, local inference means lower per-query cost and no dependency on a vendor's API uptime or pricing changes. Apple gets a fast answer to on-device AI demand; Meta gets distribution on hundreds of millions of Macs and another data point for its open-weight bet against closed labs.
The daily signal, curated. Get it in your inbox.
Subscribe on LinkedIn →