DiffusionGemma: 4x Faster Text Generation
DiffusionGemma is an experimental model that delivers 4x faster text generation.
DiffusionGemma is an experimental model built on text diffusion. Unlike traditional autoregressive large language models (LLMs), it generates text in blocks simultaneously, achieving up to 4x faster text generation on dedicated GPUs. This innovation supports the development of speed-critical interactive local workflows.