You can use generated results for commercial purposes. You own all rights to the content you create. The only restriction: make sure your prompt does not include copyrighted third-party material. You are responsible for respecting the rights to any input data.

This is a multimodal embedding model: it converts text, documents, and images into vectors, rather than generating content. It is suitable for semantic search, RAG, clustering, deduplication, ranking, and image search. Store vectors in Qdrant or pgvector and search by proximity.

None: you get a numerical vector as output, not text. Our data indicates a technical response limit of 43,253 tokens, but this refers to the API response format. For generating articles and emails, choose any generative model in BotHub.

A separate reasoning mode is not indicated in our data, and it is not needed for embeddings: the model does not build a chain of thought, but encodes the meaning of a fragment into a vector. Logic on top of the results is usually defined by your pipeline — ranking, filters, and a generative model at the final step.

Function calling and strict JSON schemas are not indicated in our data — these are tools for generative models. However, the response itself comes as a structured JSON with an array of vectors via the unified OpenAI-compatible BotHub API, so integrating the model into an existing pipeline is easy.

Images and documents — yes, that is exactly its profile: the VL model encodes images and text fragments into a single vector space, so you can search for images by description and vice versa. Audio and video input is not indicated in our data. Context is 131,072 tokens.

We have no data on languages, so we cannot make any promises — the quality of embeddings in your domain can only be verified by measurement. Collect a small set of queries and correct answers, calculate recall, and compare with another model in BotHub without changing the integration.