Google's DiffusionGemma proves you don't need to train from scratch to build a text diffusion model

Instead of training a new model from scratch, Google DeepMind retrofitted Gemma 4 into a diffusion model using less than 10 percent of the original training budget. DiffusionGemma generates 256 tokens in parallel instead of one at a time, hitting about 1,500 tokens per second. Quality still trails the original autoregressive model in benchmarks, especially o

The top 3

1. Top Companies Shaping the Future of AI: A look at the leading tech giants whose investments and innovations are setting the pace for artificial intelligence development.
2. Most Impactful Generative AI Models to Date: Exploring the groundbreaking generative AI models that have transformed creative industries and data synthesis.
3. Key Innovations in AI Model Training Efficiency: Examining the most critical advancements that have made AI model development faster, more accessible, and less resource-intensive.

Sources

• The Decoder

Open the full topic