Back to AIBriefs
AnalysisAI Models

Community releases Gemma4-26B-A4B and 31B-QAT finetunes with MTP

hauhau901 released uncensored balanced versions of Gemma4-26B-A4B and 31B-QAT with Multi-Token Prediction, boosting speed by 35% and 53% respectively. The author claims nearly 20M HuggingFace downloads and 5K Discord members.

·
18 hours ago