{"id":8892,"date":"2026-10-02T07:10:06","date_gmt":"2026-10-01T22:10:06","guid":{"rendered":"https:\/\/localmodelwatch.tsuchitsuchi.com\/2026\/10\/02\/nvidia-releases-pixelumm\/"},"modified":"2026-10-02T07:10:06","modified_gmt":"2026-10-01T22:10:06","slug":"nvidia-releases-pixelumm","status":"publish","type":"post","link":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/10\/02\/nvidia-releases-pixelumm\/","title":{"rendered":"PixelUMM Multimodal Model: 24GB+ VRAM"},"content":{"rendered":"

Sample outputs are available on the model card<\/a>.<\/em><\/p>\n

<\/p>\n

At a Glance<\/h2>\n
\n\n\n\n\n\n\n\n\n\n
Item<\/th>\nValue<\/th>\n<\/tr>\n<\/thead>\n
Repository<\/td>\nnvidia\/PixelUMM<\/a><\/td>\n<\/tr>\n
Publisher guide<\/td>\nNVIDIA: models and licenses<\/a><\/td>\n<\/tr>\n
Published<\/td>\n2026-10-02<\/td>\n<\/tr>\n
License<\/td>\nnvidia-one-way-noncommercial-license<\/td>\n<\/tr>\n
Source type<\/td>\nPrimary source (the publisher itself)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Values determined by this site’s code when the information was collected. Dates are JST.<\/em><\/p>\n

<\/p>\n

Overview<\/h2>\n

NVIDIA has released PixelUMM, a multimodal model that integrates the mutual understanding and generation of text, images, and video into a single model. This model supports a wide range of tasks, including text-to-image generation, image-to-text generation, video-text-to-text understanding, and video generation.<\/p>\n

The standout feature is its adoption of an “encoder-free” structure, where RGB images are treated directly as pixel patches rather than obtaining embeddings from separate pretrained vision encoders as in traditional models. This enables both image and video understanding and generation tasks to be processed directly in pixel space while sharing a common Transformer-based representation.<\/p>\n

Specifications<\/h2>\n

The specifications and details of PixelUMM as described in the model card and public materials are as follows:<\/p>\n