Fact · Voice and AI
3billion parameters
Apple's 2025 foundation models update describes a compact on-device model of approximately 3 billion parameters, optimized to run efficiently on Apple silicon.
Checked against its source on October 17, 2026
In the source's own words
“optimized to run efficiently on Apple silicon, and include a compact, approximately 3-billion-parameter model, alongside a mixture-of-experts server-based model”
Updates to Apple's On-Device and Server Foundation Language Models, Apple Machine Learning Research, 2025, Introduction.
Open the original- What it measures
- Apple describes two models: a compact on-device model and a server model with a mixture-of-experts design that runs in Apple's Private Cloud Compute. The post does not say which dictation features use the on-device model.
- What it does not show
- The figure is approximate, and the post does not say which apps or dictation features call this model. It describes the model that powers Apple Intelligence features, not the speech recognizer that turns sound into words.