UNIVERSAL MULTIMODAL EMBEDDINGQWEN3.5 · 9B

Meaning has
no file type.

Project language, imagery, video, charts, and visual documents into one shared geometry—then watch meaning arrange itself.

4,096native dimensions
80.6MMEB-v2 average
190MMEB-v3 tasks
CURATED EXPEDITIONSChoose a path into the model
01 / RETRIEVAL UNIVERSE

Ask in one medium.
Discover in another.

Search the curated field or add a small collection of your own. Every result is scored in the same normalized vector space.

QUERY COMPOSER
⌘ ↵

Visual media and text become one interleaved query.

Build your candidate collectionoptional
Curated universe10 mixed-media candidates
EMBEDDING BUDGET1,024D

Smaller vectors trade storage for retrieval fidelity.

MODEL READYtencent / WeMM-Embedding-9B
AWAITING QUERY

Your semantic field is quiet.

Compose a query or choose a curated expedition. The first run maps the multimodal corpus; later runs reuse its cached vectors.

Llama 4 model card screenshot Historical budget infographic Road safety assessment page
4096
ONE MODEL · MANY RESOLUTIONS

Fold the vector.
Keep the meaning.

WeMM emits a 4,096-dimensional embedding with Matryoshka structure. Store fewer leading dimensions when latency or memory matters, and inspect the tradeoff here before choosing.