NVIDIA's Diffusion-Based Nemotron Model

NVIDIA released Nemotron-TwoTower-30B-A3B-Base-BF16, a diffusion-based language model using a two-tower architecture with parallel token generation, built on the Nemotron 3 Nano 30B-A3B backbone.

One continuously updated timeline instead of dozens of separate articles. New developments are appended as the story evolves.

1. ReleaseJun 25, 2026, 08:34 AM 66%

   NVIDIA releases Nemotron-TwoTower-30B-A3B-Base-BF16, a diffusion-based language model with parallel token generation.

   NVIDIA released Nemotron-TwoTower-30B-A3B-Base-BF16, a diffusion-based language model using a two-tower architecture with parallel token generation, built on the Nemotron 3 Nano 30B-A3B backbone.

   Read the full story →