Multi Qa Mpnet Base Dot V1
Neural Network
Multi Qa Mpnet Base Dot V1 API is an embedding model that converts text into 768-dimensional vectors for semantic search and QA tasks.
Specifications
- Input
- text
- Output
- vectors (embeddings)
- Context size
- 512 tokens
- Vector dimensions
- 768
Prompt cost
(per 1M tokens)
How it works Multi Qa Mpnet Base Dot V1?
Multi Qa Mpnet Base Dot V1 encodes sentences and short paragraphs into a dense 768-dimensional vector, designed for answer retrieval, semantic search, and text similarity assessment. Via BotHub, the model is available via a unified OpenAI-compatible API: no VPN or foreign cards required, pay-as-you-go in rubles, and AES-GCM encryption. Corporate teams get contracts, invoices, and an admin panel with limits. Vectors can be stored in Qdrant, pgvector, or other vector databases, enabling RAG search across knowledge bases and documentation, finding similar support tickets, clustering reviews and logs, ranking answers in FAQ bots, and deduplicating product cards.Frequently asked questions about Multi Qa Mpnet Base Dot V1
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.
It is an encoder: it turns text into a numerical vector. These vectors are used to build semantic search, question answering, RAG pipelines, deduplication, and document clustering. The window is 512 tokens, so split long texts into appropriately sized chunks beforehand.
There is no text or image generation here — the output is an embedding, a vector of 768 numbers, which the model is optimized to compare using the dot product, as hinted by 'dot' in the name. The vector arrives as a standard JSON response from the endpoint and is stored in Qdrant, pgvector, or another vector database.
Reasoning mode is not indicated in our data, and by nature, this is an encoder model: it encodes text into a vector in a single pass, rather than reasoning or writing an answer. Add logic and explanations using a language model working on top of the retrieved fragments.
Function calling and structured output are not indicated in our data. Embedding endpoint requests and responses are in JSON, but the vector itself is not a tool for function calling. If you need tool calls, place a text model alongside it; the API in BotHub is unified and OpenAI-compatible.
Accepting files other than text is not indicated in our data: you provide text and receive a vector. For multimodal scenarios, select a model in the same window that supports such formats — you won't have to rewrite the integration, as the key and request structure remain the same.
Languages are not specified in our data, so you should check the quality on your own corpus: collect a few dozen question-document pairs and see how the output ranks by vector proximity. If the metrics are unsatisfactory, it is easy to switch to another embedding model in BotHub.