Tech advancements show SLM speed gains and Apple MacBook Pro redesign

Recent technical developments highlight advancements in both hardware design and software optimization for artificial intelligence. Tests conducted on an M2 MacBook Air using the Qwen2.5-0.5B-Instruct model demonstrate that length-based batch processing significantly improves the performance of Small Language Models (SLMs). By sorting sequences by length to reduce padding overhead, processing speeds increased from 4.2 tokens per second in individual processing to 7.5 tokens per second in batch mode, effectively doubling efficiency without sacrificing accuracy.

In hardware news, Apple is reportedly working on a redesign of the MacBook Pro to increase portability. According to Bloomberg, the company is shrinking certain internal components and reorganizing the internal layout to make the upcoming 14-inch and 16-inch models noticeably lighter. This redesign may coincide with other potential upgrades, such as OLED displays and thinner chassis, marking a significant departure from the design language established in 2021.