All MiniLM L6 V2
Neural Network
All MiniLM L6 V2 is an embedding model that converts sentences and short paragraphs into 384-dimensional vectors for semantic search.
Specifications
- Input
- text
- Output
- vectors (embeddings)
- Context size
- 512 tokens
- Vector dimensions
- 384
Prompt cost
(per 1M tokens)
How it works All MiniLM L6 V2?
All MiniLM L6 V2 maps sentences and short paragraphs to a dense 384-dimensional vector space, providing compact representations for semantic search, clustering, and ranking. The compact dimensionality reduces index size and speeds up nearest neighbor search in Qdrant, pgvector, FAISS, and other vector databases. In BotHub, the model connects via a unified OpenAI-compatible API alongside 250+ other models. Access is available without a VPN or foreign card, with payment in rubles based on token usage. Use it for RAG pipelines over knowledge bases, ticket deduplication and clustering, semantic search in documentation, product recommendation, and short text classification.Frequently asked questions about All MiniLM L6 V2
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 an embedding model from sentence-transformers: it turns text into a numerical vector used to measure semantic similarity. It is suitable for semantic search, RAG-based document search, clustering and deduplication of records, classification of inquiries, and building recommendations based on meaning rather than exact word matches.
The output is not text, but an array of numbers — a vector representation of the input fragment. The context window is 512 tokens, so split long documents into chunks beforehand. Store the resulting vectors in Qdrant, pgvector, or another vector storage and search by cosine similarity.
Step-by-step reasoning is not noted in our data for this model, and its task does not imply it: an embedder encodes text into a vector in a single pass. If your pipeline requires reasoning and answer formulation, place a text model alongside it — in BotHub, they are available in the same interface and via the same API.
Function calling and structured output are not noted in our data for this model. It is accessed via the unified OpenAI-compatible BotHub API, so the response comes as a standard JSON with an array of vectors, which is convenient to pass directly to your indexing system.
Acceptance of data types other than text is not noted in our data — work with text fragments. If you need to vectorize the content of images or call recordings, first obtain a text description or transcript using a suitable model, and then pass the result here for indexing.
We do not have data on the languages the model was trained on, so it is more reliable to check the quality on your own corpus: collect a small set of queries with known correct answers and compare the output. Access to BotHub works from Russia without a VPN or foreign card, with payment in rubles.