AI Matters
Why it matters
Google releases DiffusionGemma to speed up text generation
The Decoder Google's DiffusionGemma proves you don't need to train from scratch to build a text diffusion model
Google DeepMind converted its Gemma 4 model into a text diffusion model called DiffusionGemma, which generates up to 1,500 tokens per second by processing text in parallel.
Why it matters
Generating text in parallel instead of word-by-word could dramatically lower the computing costs and time required to run AI models.
Latest brief · Aug 9, 2026