DiffusionGemma
#3 in Text-to-Imagegoogle · since 2026-06-10 · 22× · last seen Jun 29, 2026
69
Momentum
DiffusionGemma is an experimental open-source language model from Google with 26 billion parameters (Mixture of Experts) that is based on diffusion-based techniques instead of linear generation. It was released under Apache 2.0 license and can perform text generation up to 4 times faster than other Gemma 4 variants. The model is natively supported in vLLM.
Momentum trend
04.04.03.07.
Features
| License | Apache 2.0 (additionally Gemma Terms of Use and Gemma Prohibited Use Policy) |
| Max Video Length | Max. 60 seconds of video input (processed at 1 frame per second) |
| Release Date | June 10, 2026 |