# Google Releases Open-Source DiffusionGemma 26B Model

> The new 26B parameter model from DeepMind uses a diffusion-based architecture, a technique more common in image generation, to produce text.

Published by The Open Weights on Jun 9, 2026. Canonical: https://theopenweights.com/news/diffusiongemma-26b-a4b-h4ke

## Key facts

- Company: Google DeepMind
- Model: DiffusionGemma
- Version: 26B-A4B-it
- Category: Text / LLM
- Modalities: Text / LLM, Vision-Language
- License: Apache 2.0 (Open weights, commercial use allowed)
- Parameters: 26B (mixture of experts)
- Active parameters: 3.8B
- Context window: 262K tokens
- Architecture: DiffusionGemmaForBlockDiffusion
- Precision: BF16
- Weights size: 51.6 GB
- Languages: 35+ languages
- Significance: notable
- Published: 2026-06-09
- Last verified: 2026-06-28
- Hugging Face: https://huggingface.co/google/diffusiongemma-26B-A4B-it
- Canonical URL: https://theopenweights.com/news/diffusiongemma-26b-a4b-h4ke

Google DeepMind has released DiffusionGemma, a new family of open-source language models that explores a novel architecture for generating text. The first release is an instruction-tuned, 26-billion-parameter model, signaling Google's continued investment in both the Gemma family and alternative AI architectures.

## A Different Approach to Generation

Unlike most large language models like GPT-4 or Llama 3, which are autoregressive and generate text one token at a time, DiffusionGemma employs a diffusion-based method. This technique, widely used in popular image generation models, starts with a complete, noisy sequence of text and iteratively refines it into a coherent output. This marks a notable departure from the standard Transformer architecture that has dominated natural language processing for years.

The newly released model is a Mixture-of-Experts (MoE) variant, which means that while it has 26 billion total parameters, only 4 billion are active for any given input. This design makes inference more computationally efficient than a dense model of the same size. According to its official model card, the [DiffusionGemma 26B-A4B-it model](https://huggingface.co/google/diffusiongemma-26B-A4B-it) is designed for a variety of image-to-text and text-to-text tasks.

By open-sourcing DiffusionGemma under a permissive Apache 2.0 license, Google is enabling the wider AI community to experiment with this less-common approach. The exploration of non-autoregressive models could lead to new discoveries in text generation, potentially offering different trade-offs in speed, coherence, and controllability compared to today's leading models.

## Get the model

- [Hugging Face](https://huggingface.co/google/diffusiongemma-26B-A4B-it)

## Sources

- [google/diffusiongemma-26B-A4B-it](https://huggingface.co/google/diffusiongemma-26B-A4B-it) — Hugging Face, Jun 9, 2026

---
Source: The Open Weights (https://theopenweights.com/). Aggregated and written by Claude, curated by humans. Cite as: "Google Releases Open-Source DiffusionGemma 26B Model", The Open Weights, Jun 9, 2026, https://theopenweights.com/news/diffusiongemma-26b-a4b-h4ke