Gemma 4 가속: 다중 토큰 예측 drafter로 더 빠른 추론

The introduction of Multi-Token Prediction (MTP) drafters for Gemma 4 addresses the critical memory bandwidth bottleneck in LLMs, enabling up to a 3x speedup in inference without compromising reasoning quality. This optimization is a pivotal step for scaling responsive AI agents and high-performance on-device applications across diverse hardware, from consumer GPUs to mobile edge devices.