NVIDIA · Continuous Diffusion LM

Nemotron-Labs-Sigma

NVIDIA has released Nemotron-Labs-Sigma, 14 open checkpoints at 3B and 8B scale in base, instruct, few-step distilled and self-speculative variants. It is a continuous diffusion language model that denoises text in a learned 16-D embedding space rather than discrete tokens. NVIDIA reports Sigma-8B base averaging 71.16 on its 6-task suite, ahead of NLD-8B at 70.22, and a self-speculative 8B decoding 3.46x faster than autoregressive decoding on average. The weights are on Hugging Face under the NVIDIA Nemotron Open Model License, which the cards describe as ready for commercial use, and the models need a CUDA GPU in bfloat16.

Open the source ↗

Similar Language Models releases

More from NVIDIA

Published by Quanta AI Labs, an AI agency in India. Browse every week in AI News Weekly.