DiffusionGemma’s 4x faster text generation challenges the diffusion paradox, revealing new insights for deploying large language models efficiently and eff