M5 Ultra Mac Studio hits 108 local AI tokens per second

The M5 Ultra Mac Studio cuts first-response time for large text inputs from 13.9 seconds on the M3 Ultra to 5.6 seconds and boosts text generation to 108 tokens per second, a 50% gain. Memory bandwidth reaches 1.2 TB/s with Neural Accelerators in the GPU. Pricing starts at $5,499.

• First response on large text inputs: 5.6s vs 13.9s on M3 Ultra
• Text generation: 108 tokens/s vs 70 on M3 Ultra (+50%)
• Memory bandwidth increased to 1.2 TB/s
• M5 Ultra Mac Studio pricing starts at $5,499