E5 Base V2
Neural Network
E5 Base V2 is an embedding model that encodes texts into 768-dimensional vectors for semantic search and RAG.
Specifications
- Input
- text
- Output
- vectors (embeddings)
- Context size
- 512 tokens
- Vector dimensions
- 768
Prompt cost
(per 1M tokens)
How it works E5 Base V2?
E5 Base V2 converts individual sentences and entire paragraphs into dense 768-dimensional vectors, balancing semantic representation quality and index size. The resulting vectors can be directly stored in Qdrant, pgvector, and other vector databases. In BotHub, the model is available via a unified OpenAI-compatible API: access from Russia without VPN or foreign cards, pay in rubles per token, encrypted requests, and switching to another embedder requires no integration rewrite. Use it for semantic knowledge base search, building RAG pipelines, document clustering and deduplication, text similarity assessment, and ranking similar materials in search results.Frequently asked questions about E5 Base 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.
The model from intfloat turns text into embeddings — numerical vectors used to calculate semantic similarity. Suitable for semantic search, RAG pipelines, document clustering and deduplication, classification, and finding similar materials. Context is 512 tokens, so split long texts into chunks beforehand.
The output is not text, but a fixed-dimension vector — 768 values for the base version, an array of floating-point numbers in the JSON response. These vectors are directly written to Qdrant, pgvector, or other vector storage and compared by cosine similarity.
Reasoning mode is not indicated in our data, and it is not expected for an embedding model: e5-base-v2 does not conduct dialogue or write answers; it encodes the input text into a vector in a single pass. Your system builds logic on top of the search results.
Function calling and structured output schemas are not indicated in our data — these are tools for generative models. Access is via the unified BotHub OpenAI-compatible API at the embeddings endpoint, and the result is returned in JSON with an array of vectors ready for database insertion.
Support for other modalities is not indicated in our data: input text — a string or an array of strings for batch processing, with a 512-token limit per chunk. If you need vectors for other data types, you can switch to another model in the BotHub window without rewriting the integration.
Language support is not specified in our data, so we will not promise quality for a specific language — run your own sample of queries and documents to measure search accuracy. Access to the model from Russia works without VPN or foreign cards, with payment via Russian cards.