Apple's M6 and M5 Pro show massive gains in local AI performance

Apple’s latest silicon is showing how quickly on-device AI performance is scaling.
The M6 reportedly delivers up to 13.5x faster LLM prompt processing than the M1 and around 4.8x the performance of the M4.
Meanwhile, the M5 Pro reaches up to 4x the LLM performance of the M4 Pro.
Thunderbolt 5 also allows multiple Mac minis to be clustered together, making it possible to run significantly larger AI models entirely on-device.
For $AAPL, the AI strategy is increasingly about turning Macs into local inference machines rather than relying entirely on cloud compute.