Skip to content
AI Model Radar

diffusiongemma 26B A4B it

google/diffusiongemma-26B-A4B-it
Open Source · chat · open-weights
Open Alert me on changes
Intelligence
9.5
Context
262.1K
Max output
32.8K
Weights
Open
API $/1M
$0.04 / $0.22
Modalities
text · image
Released
09 Jun 2026
Intelligence Index via Artificial Analysis · 0–100, higher is better
License: apache-2.0 · google/diffusiongemma-26B-A4B-it
Download image Share on X Share on LinkedIn
AI summary
● machine-written

DiffusionGemma 26B A4B: Google's discrete diffusion model on Gemma 4 architecture

DiffusionGemma 26B A4B is a generative model built by Google DeepMind based on the 26B A4B Mixture-of-Experts Gemma 4 architecture, generating tokens using discrete diffusion. It supports text and image input/output with a 256K token context window. The model is available as open weights under Apache 2.0 license.

What's new
  • Based on 26B A4B Gemma 4 MoE architecture with selective token activation
  • Uses discrete diffusion for token generation
  • Supports multimodal inputs: text, images, and video
  • 256K token context window
Best for
Text and image generation tasksMultimodal applications requiring discrete diffusionOpen-source deployment scenarios
Sources

Source: https://huggingface.co/google/diffusiongemma-26B-A4B-it