DiffusionGemma 26B A4B
ResearchOpen weightsText26B total / 4B active (MoE)
DiffusionGemma family model by Google.
About the DiffusionGemma family
An experimental open Gemma variant using text diffusion instead of autoregression — generating whole passages in parallel refinement steps.
All DiffusionGemma modelsHow this one is configured
This configuration mixture-of-experts with ~26B total / ~4B active parameters per token.