Text Embedding 3 Large
Neural Network
text-embedding-3-large is an OpenAI embedding model for semantic search, RAG, and text clustering.
Specifications
- Input
- text
- Output
- vectors (embeddings)
- Context size
- 8 192 tokens
- Vector dimensions
- 3 072
Prompt cost
(per 1M tokens)
How it works Text Embedding 3 Large?
text-embedding-3-large is an OpenAI embedding model that converts text into a numerical vector: the closer the vectors, the closer the meaning of the source fragments. The model works with texts in different languages, and vector proximity is calculated using cosine similarity — the basis for semantic search rather than keyword matching. In BotHub, it connects via a unified OpenAI-compatible API without VPN or foreign cards, with pay-as-you-go payment via Russian cards and AES-GCM encryption. Store ready-made vectors in Qdrant or pgvector: semantic search over documentation, RAG over knowledge bases, clustering and deduplication of requests, recommendations of similar materials, and ticket classification. For teams: contracts, invoices, and limits.Frequently asked questions about Text Embedding 3 Large
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.
The OpenAI model converts text into numerical vectors, which can be used to calculate semantic similarity. It is suitable for semantic search, RAG systems over your knowledge base, document clustering and deduplication, request classification, and recommendations. Context: 4095 tokens per fragment.
The output is not text, but an array of floating-point numbers — a vector representation of the input fragment. This vector can be directly stored in Qdrant, pgvector, Milvus, or any other vector storage, and proximity is calculated using cosine distance.
This is an embedding model, not a chat model: it does not conduct dialogue or build reasoning chains, but encodes the meaning of text into a vector. Build reasoning logic at the pipeline level — vector search, followed by an answer from a generative model via the same BotHub API.
Our data does not indicate this, and by its nature, an embedding model returns a vector, not an arbitrary structure. However, the API response comes in a standard OpenAI-compatible JSON format, so integrating it into existing code is straightforward.
Our data does not indicate support for files other than text — the model is designed for text fragments. If you need to work with other modalities, multimodal models are available in BotHub via the same OpenAI-compatible API, and switching will not require rewriting the integration.
The OpenAI embedding line was trained on multilingual data, and Russian is included: semantically similar Russian phrases receive similar vectors. Before launching, check the quality on your own sample — the result depends on the domain, fragment length, and document chunking method.