Back to AIBriefs
AnalysisAI Models

Gemma 4 QAT uncensored fine-tunes get MTP speed boost

Gemma 4 QAT uncensored fine-tunes get MTP speed boostView on Reddit

Community fine-tunes of Gemma 4 (26B-A4B and 31B-QAT) add Multi-Turn Prediction (MTP), boosting inference speed by 35% and 53% respectively. Models are available on Hugging Face.

·
Jun 25, 12:13 AM
Gemma 4 QAT uncensored fine-tunes get MTP speed boost — AIBriefs