by AIDC-AI
Open multimodal embedding models from AIDC-AI that map text, images, video and documents into a shared vector space.
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.
Ovis Embedding is a family of multimodal embedding models from AIDC-AI, published in the ATH-MaaS collection on Hugging Face. The models encode text, images, video and documents into a single shared embedding space, so they can be used for cross-modal retrieval and search.
According to a YouTube creator summary, the family includes several sizes, and the Omni 3B variant additionally handles audio. The same source describes the models as state-of-the-art among models of similar size and says they are released under the Apache 2.0 license. These claims come from a secondary source and were not checked against the model cards.
The weights are distributed through Hugging Face, so users can run them locally or on their own infrastructure. Details such as exact variants, benchmark numbers, embedding dimensions and release date were not available in the provided material.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Multimodal embedding models mapping text, images, video and documents (Omni 3B also audio) into one space; state-of-the-art among similar sizes, Apache 2 license.
Closest neighbours by meaning, not just by category.
Model · Google
Google's small open embedding model that maps text, code, images, video and audio into a shared embedding space.
Model · Perplexity
Open-weights vision-language embedding models from Perplexity (9B and 600M) for text, image and document retrieval.
Model · NVIDIA
NVIDIA family of multilingual text embedding models (8B and 1B) for retrieval, RAG and coding agents.
Model · Baidu
Baidu's open-weights vision-language OCR model for parsing very long, multi-page documents.
Model
Vision Transformer foundation model for sheet music, pre-trained with masked autoencoding on 9.7M IMSLP pages.
Model · DeepSeek
Experimental vision-enabled version of DeepSeek V4 Flash with image understanding; open weights under MIT, now retired on the API.