Sentence Transformers: multi-qa-mpnet-base-dot-v1

Sentence Transformers: multi-qa-mpnet-base-dot-v1 is an embedding model from Sentence Transformers, a provider known for its sentence-level representation models. It sits in the embeddings category, which means it does not generate text. Instead, it converts text into vectors that you can store, compare and search over. The model was released by the provider on 2025-11-18.

The model takes text as input and returns embeddings as output. It maps sentences and short paragraphs into a 768-dimensional dense vector space, and it is optimised for question-and-answer retrieval, semantic search and similarity scoring across diverse content. The context window is 512 tokens, so each request should contain a single sentence or a short paragraph rather than a long document. The maximum output is 460, which in this case relates to the embedding output rather than generated text.

Because the model is tuned for question-and-answer retrieval, the most direct use case is building a retrieval layer over a document collection. You embed your passages once, embed the user's question at query time, and rank passages by similarity. The same property makes it useful for semantic search, where you want matches based on meaning rather than exact keyword overlap, and for similarity scoring between two pieces of text, such as deduplication or grouping related items. It also fits clustering and classification pipelines that need fixed-length vector representations of short text.

You call it through api14's OpenAI-compatible API using the model id "sentence-transformers/multi-qa-mpnet-base-dot-v1". Because the API follows the OpenAI format, you can point an existing OpenAI client at the api14 base URL and send your text to the embeddings endpoint. The model supports no additional request parameters, so keep requests simple and rely on the defaults. Send the text you want embedded, then read the returned vector from the response and store it in your vector database or compare it directly against other vectors.

The main constraint to keep in mind is the 512 token context window. Long documents will need to be split into sentences or short paragraphs before embedding, and you should plan your chunking strategy accordingly. If your task requires generating text, summarising or answering questions directly rather than producing vectors, an embeddings model is not the right fit and you should choose a text generation model instead.