Multi-codebook diffusion LLM from Shanghai AI Laboratory for visual understanding, image generation and image editing.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
Intern Lumina U2 is a diffusion large language model from the Intern Lumina U2 Team at Shanghai AI Laboratory (InternLM). It uses discrete multi-codebook (VQ-level) visual tokens in a single unified architecture, with parallel positions and sequential codebooks. The project page describes it as covering text question answering, image understanding, image editing, image generation, video understanding and 3D understanding.
The project page reports benchmark results against InternVL-U, LLaDA2.0-Uni and LLaDA-o on documents and charts (ChartQA, CharXiv-DQ), perception (HallusionBench, MME-C, RealWorldQA), reasoning (MMMU, MMMU-Pro, MathVista, MathVision, DynaMath), generation (GenEval, DPG-Bench, TIIF, WISE) and editing (ImgEdit). Video and 3D results are compared against specialist models such as VideoLLaMA2-7B, LLaDA-V, PointLLM-7B and ShapeLLM-7B. The scores are for the released Ascend-trained checkpoint. These are vendor-reported numbers.
According to a YouTube creator, code instructions have been released but weights were not yet on Hugging Face at the time of that video. The page text provided does not state a license, so availability of weights and licensing terms should be verified against the official resources. A technical report is cited as arXiv:2610.20818.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Open-source diffusion large language model that generates and edits images and also has strong image understanding, such as identifying scans, paintings and ER diagrams. Code instructions are out but weights are not yet on Hugging Face.
Closest neighbours by meaning, not just by category.
Model · Princeton
Fully open 3B text-to-image diffusion model from Princeton with released weights, training code, and data pipeline.
Model · Linum AI
2B-parameter pixel-space text-to-image diffusion model from Linum AI, trained from scratch for about $21K; Apache 2.0.
Model · inclusionAI
Open-weights 6B unified text-to-image and instruction-based image editing model family from inclusionAI.
Model · ByteDance
Diffusion-based vision-language model from PKU MSALab and ByteDance that captions multiple image regions in parallel.
Model · inclusionAI
Open-weights 560B hybrid-linear MoE reasoning model (25B active, 1M context) from inclusionAI, MIT licensed.
Model · Ant Group
Ant Group's multimodal mixture-of-experts model (560B total, 25B active) in the Ling family, served via API.
