With Apple’s introduction of the M5 Max, as detailed by Alex Ziskind in his YouTube video “Apple’s New M5 Max Changes the Local AI Story” (published on March 9, 2026), there’s a palpable shift in how local AI is perceived, particularly in terms of processing power and speed. The M5 Max MacBook Pro showcases a new GPU architecture equipped with neural accelerators within each GPU core, claiming significant performance enhancements over its predecessor, the M4 Max. Ziskind examines these claims through various benchmark tests, focusing on multi-core processing capabilities crucial for software development and local AI applications.
The M5 Max boasts impressive numbers, featuring a 40-core GPU promising four times the peak GPU compute for AI compared to previous models, alongside a memory bandwidth of 614 GB/s. These changes suggest a substantial leap, especially evident in Ziskind’s tests where the Mandelroad algorithm, designed to test multi-core performance, demonstrated that the M5 Max significantly outpaced the M4 Max and wasn’t far behind the M3 Ultra Max Studio.
One of the video’s highlights was the exploration of local LLM processing performance, where the M5 Max showed notable improvements over the M4 Max in token generation speed—a key aspect in AI computations. This was attributed to advances in memory bandwidth and SSD speeds, with the M5 Max’s SSD read speeds nearly doubling those of its predecessors. This enhancement was particularly significant for large model loading and prompt processing, a fact Ziskind underscored using the LM Studio’s Quen 3.5 AI model as a test case.
However, the M5 Max wasn’t without its limitations. Despite improved token generation times, the margin over the M3 Ultra was narrower than anticipated, suggesting that while the M5 Max is powerful, the advantages may vary depending on the specifics of AI models and use cases. Ziskind’s test with the GPTOSS 12B model highlighted that although the M5 Max outperformed the M4 Max, it didn’t leap as dramatically as perhaps initially expected.
Still, the M5 Max marks a fascinating progression in local AI capabilities. Its performance in prompt processing, thanks to its built-in neural accelerators, offers a glimpse into the future potential of desktop-like power in a portable form. As Ziskind concludes, the M5 Max’s advancements represent a significant step forward, capitalizing on Apple’s robust hardware ecosystem.
Overall, the video paints a compelling picture of Apple’s technological innovations and the potential these advancements hold for the everyday computing experience, particularly for developers and AI enthusiasts. While some skeptics might question the incremental benefits in certain areas, the general consensus is appreciation for Apple’s commitment to pushing the boundaries of portable computing power.