Apple brings on-device generative AI to iOS 18.4 beta
Apple confirmed its biggest generative AI integration yet in the iOS 18.4 beta, running models entirely on-device via the NPUs in A18 and M4 chips instead of the cloud. Internal documents put throughput at 45 tokens per second on the A18 Pro, a 30% gain over the previous version.
- Model runs at 45 tokens/s on A18 Pro, 30% faster than before
- Private Compute framework keeps user data off external servers
- New API suite opens on-device AI to third-party developers
- Final iOS 18.4 release expected in late spring
Read next
Software