Gemma 4 26B A4B
Google
Gemma 4 26B A4B IT is an instruction-tuned Mixture-of-Experts (MoE) model from Google DeepMind. Despite 25.2B total parameters, only 3.8B activate per token during inference — delivering near-31B quality at...
Parameters
25.8B
Context
262144
Modality
text+image+video->text
License
proprietary
Open Gemma 4 26B A4B in AIniverse
Compare versions, read real ratings, save to your stack.
Open in App