AI Matters

Why it matters

Google introduces DiffusionGemma to generate text in parallel

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

Google DeepMind retrofitted its Gemma 4 model into a text diffusion model, allowing it to generate up to 1,500 tokens per second by processing them in parallel.

Why it matters

Generating entire blocks of text at once rather than word-by-word could significantly speed up AI assistants and reduce computing costs.