Google's open-source diffusion language model generates 256 tokens in parallel and self-corrects, hitting 4x speed on one GPU at a cost to quality.

DiffusionGemma generates text up to 4x faster than traditional models by producing entire blocks simultaneously, achieving roughly 1,479 tokens per second.

The new DiffusionGemma open model generates text in parallel — not one token at a time — and is optimized to run on the NVIDIA RTX PRO platform, NVIDIA DGX Spark systems and…