option
Home
Flash News
Content
JuanTaylor
JuanTaylor
July 3, 2026

NVIDIA released the Nemotron-Labs-TwoTower discrete diffusion language model on July 2, featuring a 60B parameter dual-tower architecture with two 30B networks. It uses parallel text generation via diffusion, boosting throughput by 2.42x while retaining 98.7% of original capabilities. The model is open-sourced on Huggingface and requires two H100 or A100 80GB GPUs for full inference.

NVIDIA released the Nemotron-Labs-TwoTower discrete diffusion language model on July 2, featuring a 60B parameter dual-tower architecture with two 30B networks. It uses parallel text generation via diffusion, boosting throughput by 2.42x while retaining 98.7% of original capabilities. The model is open-sourced on Huggingface and requires two H100 or A100 80GB GPUs for full inference.
Comments (0)
0/300
OR