Back to AIBriefs
AnalysisAI Models

Community releases Gemma4 QAT models with MTP speed boosts

HauhauCS released Gemma4-26B-A4B and 31B-QAT Balanced QAT models with Multi-Token Prediction, claiming 35% and 53% speed boosts. The uncensored fine-tunes are already trending on HuggingFace with over 28,000 downloads.

·
Jun 24, 11:43 PM