NVIDIA releases Dory1.0-4B, a 4B looped language model with adjustable inference compute
Overview
NVIDIA releases NVIDIA-Nemotron-Labs-Dory1.0-4B-MXFP8, a 4B-parameter looped language model with up to 256K context, on Hugging Face under the OpenMDW-1.1 license.
The model applies a shared recurrent block several times, so users can change the loop count to trade inference compute for accuracy without retraining. The model card reports benchmark scores at 1, 2, and 4 loops, and describes a last-loop KV caching mode that keeps the cache at one-loop size. The license states the model is ready for commercial or non-commercial use.
Written by AI from the articles below · updated Oct 16, 8:38 PM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- NVIDIA · new models on Hugging FaceOfficialNVIDIA releases Dory1.0-4B, a 4B looped language model with elastic inference compute
AINVIDIA has released NVIDIA-Nemotron-Labs-Dory1.0-4B-MXFP8, a 4B-parameter looped language model with up to 256K context under the OpenMDW-1.1 license. The model applies a shared recurrent block multiple times, letting users change loop count to trade inference compute for accuracy without retraining. The card lists a last-loop KV caching mode that keeps the cache at one-loop size, and reports benchmark scores at 1, 2, and 4 loops.
Heat trend
Not enough continuous observations to show a trend yet.