Diffusiongemma
Google DeepMind🇺🇸 USFrontier· Jun 9, 2026
TextVisionopen weights
Open-weights multimodal model from Google DeepMind that takes text, image and video input and outputs text. It denoises 256-token blocks in parallel via diffusion instead of generating token by token, aimed at speed. The model card reports most benchmark scores below those of Gemma 4 26B A4B.
Summary of the model card
- License
- apache-2.0
- Modalities
- text, image → text
1 variant
| 26B A4B IT | 26B-A4B | HF |
Sources
HuggingFaceDeutsch
Offenes multimodales Modell von Google DeepMind, das Text, Bilder und Video verarbeitet und Text erzeugt. Es erzeugt Bloecke von 256 Tokens parallel per Diffusion statt Token fuer Token, ausgelegt auf hohe Geschwindigkeit. Laut Modelcard liegen die Benchmark-Werte meist unter denen von Gemma 4 26B A4B.
Provenance
- Detected
- Sep 27, 2026
- First source
- huggingface
- Description
- Summary of the model card
- Editorially reviewed
- —