Tasks and inputs
WeMM Embedding 4B is listed for Embeddings, Visual document retrieval. It accepts text, image, video and produces embedding. Use the publisher’s input format and preprocessing for your chosen deployment.
WeMM Embedding 4B encodes text, images, video and visual documents for multimodal similarity search.
The specification score summarizes recorded features and limits. It is not a benchmark result.
Recorded modality support shows what this model can accept and produce. An undocumented modality is shown as unknown rather than assumed to be unsupported.
Accepts
Produces
WeMM Embedding 4B encodes text, images, video and visual documents for multimodal similarity search.
Tencent’s 4B checkpoint produces 2560-dimensional normalized vectors. The documented interface accepts text, images, video, visual documents and interleaved input. Audio support should not be inferred from the multimodal name. The release uses Apache 2.0 and provides task instructions for query and document encoding.
For this version, compare retrieval quality and serving cost with the smaller checkpoint. Build a labelled query set covering exact terms, paraphrases, image questions and queries whose evidence is spread across a page. Score whether the intended results appear near the top, and inspect false matches as well as missed relevant items.
Keep the selected model, output dimension, normalization and query or document template consistent from indexing through search. Replacing another embedding model generally means rebuilding stored vectors. Measure storage and query latency for the actual collection size rather than using encoder parameter count as a search-speed estimate.
The output is an embedding, not a written answer. If a later language model explains retrieved material, evaluate that generation separately and keep its citations tied to the retrieved evidence. Hosted prices and chat-style output limits remain unknown unless a selected provider documents them.
Compare the model on the work you need to complete.
WeMM Embedding 4B is listed for Embeddings, Visual document retrieval. It accepts text, image, video and produces embedding. Use the publisher’s input format and preprocessing for your chosen deployment.
Evaluate retrieval over a labelled query set. Keep the encoder, vector dimensions and index settings consistent.
The weights are downloadable under Apache 2.0. Check runtime support and size the complete model and working context for your expected concurrency.
These records share a directory category. Check the source card, output type and deployment requirements before treating a model as a replacement.
EmbeddingGemma 2 creates shared embeddings for text, code, images, video and audio for multimodal retrieval.
Tencent
WeMM Embedding 2B encodes text, images, video and visual documents for multimodal similarity search.
Tencent
WeMM Embedding 9B encodes text, images, video and visual documents for multimodal similarity search.