GTE Base
Neural Network
GTE Base is an embedding model that converts text into 768-dimensional vectors for semantic search and clustering.
Specifications
- Input
- text
- Output
- vectors (embeddings)
- Context size
- 512 tokens
- Vector dimensions
- 768
Prompt cost
(per 1M tokens)
How it works GTE Base?
GTE Base encodes sentences and paragraphs into dense 768-dimensional vectors — compact representations for evaluating semantic similarity, search, and clustering. This dimension is convenient for vector databases like Qdrant and pgvector: indexes remain lightweight, and search is fast. In BotHub, the model is available via a unified OpenAI-compatible API, so switching to another model does not require rewriting the integration; access from Russia is available without a VPN or foreign card, with payment in rubles based on token usage. Use gte-base for RAG, documentation and knowledge base search, deduplication and clustering of requests, finding similar materials, and text classification via vectors.Frequently asked questions about GTE Base
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.
gte-base is an embedding model: it turns text into a numerical vector describing its meaning. Such vectors solve semantic search, RAG, clustering, and classification tasks, and are convenient to store in vector databases like Qdrant or pgvector. Through the unified BotHub API, the model connects to your multi-model system alongside others.
gte-base does not generate text or images. Its output is a vector representation of the input text: a set of numbers of fixed dimension that can be compared by cosine similarity, added to an index, and used for search or RAG. The response comes in a machine-readable format, convenient for code processing.
In our data, reasoning before answering is not noted for this model. gte-base is an encoder for embeddings: it immediately converts input text into a vector, without step-by-step reasoning or dialogue. If you need task analysis with explanations, there are text models available in BotHub, and switching to them does not require rewriting the integration.
In our data, function calling and structured JSON output are not noted for gte-base. Its result is a vector, not an answer according to a specified JSON schema, so function calling scenarios are not described for it. For such tasks, it is wiser to use a text model and leave gte-base for search and ranking.
In our data, the acceptance of images, audio, and video is not noted for gte-base: the model works with text and returns embeddings. If you need to process images, video, or voice, BotHub has separate models for these tasks, and switching between them does not require rewriting the integration.
The data does not contain information about the quality of gte-base's performance with the Russian language, so there is nothing to promise here: the language of prompts and output is not specified in the characteristics. It is worth checking with your own examples — take a few similar queries and compare the proximity of the resulting vectors.