Tencent's WeChat Vision team opened WeMM-Embedding, a family of universal multimodal embedding models
GitHub·low signal
Tencent/WeMM-Embedding was created 2026-08-25 and pushed the same morning, carrying topics for embedding-models, multimodal and multimodal-llm. It is described as a family of universal multimodal embedding models from the WeChat Vision Team supporting both multimodal understanding and retrieval. At 26 stars on day one it is too early to judge, and the license is non-standard rather than a recognized SPDX identifier, which matters if you plan to build retrieval on it.