09:17
2026-08-05
promptcube3.com
artificial-intelligence
Maple-Preview: 120 tok/s 20B MoE on iPhone Defies Expectations
Maple-Preview, a 20B parameter Mixture-of-Experts transformer quantized to ternary weights, achieves 120 tokens per second on an iPhone, according to a developer preview. The model, which runs on Appl…