NVIDIA's TwoTower 30B model flips token generation with block‑level diffusion
Built on the Nemotron‑3 Nano 30B‑A3B backbone, NVIDIA’s TwoTower model replaces single‑token autoregression with a block‑wise diffusion process that keeps a frozen AR tower frozen while a denoiser tower generates token groups. The diffusion approach aims to boost reasoning and planning, and the model is ready to run via Transformers, vLLM, SGLang or Docker on Hugging Face.