Tencent has released WeMM-Embedding-2B, a universal multimodal embedding model built on Qwen3.5 that accepts text, images, videos, visual documents, and interleaved inputs to return 2,048-dimensional L2-normalized embeddings.
The model supports independent encoding of any content subset and utilizes Matryoshka dimensions, where 256-dimensional embeddings retain 98.7% of full-dimensional performance on MMEB-v2.
It is available under the Apache License 2.0 with code and weights publicly released, and includes integration examples for sentence-transformers, vLLM, and SGLang.