How to Get 2x Speed on Gemma 4 with Multi-Token Prediction in llama.cpp
中文摘要
llama.cpp现已支持Gemma 4多令牌预测(MTP)，将Gemma 4推理速度提升高达2倍。此更新显著提高了效率。
English Summary
llama.cpp now supports Gemma 4 Multi-Token Prediction (MTP), enabling up to 2x speed. This update significantly boosts Gemma 4 inference efficiency.
原文节选
On June 7, 2026, PR #23398 by contributor am17an landed in llama.cpp b9549, bringing official Gemma 4 Multi-Token Prediction (MTP) support… Continue reading on Medium »