Mastodon Feed
Boosted by jsonstein@masto.deoan.org ("Jeff Sonstein"):
viticci@macstories.net ("Federico Viticci :ticciseal:") wrote:
I finally tested the M5 iPad Pro with local AI and Neural Acceleration in the GPU.
The performance improvement from the M4 in time-to-first-token is WILD.
Details and charts in my story: https://www.macstories.net/stories/ipad-pro-m5-neural-benchmarks-mlx/
