Gemma4-26B-A4B & 31B-QAT Uncensored Balanced models released with MTP speed boosts

Core summary

One sentence to understand this update

Gemma4-26B-A4B and 31B-QAT Uncensored Balanced models have been released, featuring significant MTP speed boosts of 35% and 53% respectively.

Impact & opportunity

What this could mean

Developers can leverage these optimized Gemma models to gain performance advantages in applications requiring high inference speeds, especially in contexts with fewer content restrictions.