GTE Large
Neural Network
GTE Large is an embedding model that turns sentences, paragraphs, and documents into 1024-dimensional vectors for semantic search.
Specifications
- Input
- text
- Output
- vectors (embeddings)
- Context size
- 512 tokens
- Vector dimensions
- 1 024
Prompt cost
(per 1M tokens)
How it works GTE Large?
GTE Large is an embedding model that converts sentences, paragraphs, and documents of moderate length into dense 1024-dimensional vectors. This representation is suitable for information retrieval, semantic text similarity assessment, and re-ranking. In BotHub, you access the model via a single OpenAI-compatible API: one key, and switching to another model does not require rewriting the integration. Access is available from Russia without a VPN or foreign card, payment in rubles based on actual token usage, and traffic is encrypted with AES-GCM. Typical scenarios: RAG search in Qdrant or pgvector knowledge bases, clustering and deduplication of requests, matching similar products or vacancies, and text classification.Frequently asked questions about GTE 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.
gte-large turns texts into vectors — numerical representations of meaning. It is suitable for semantic search, RAG, clustering, deduplication, and classification: semantically similar fragments end up close to each other in vector space. Store the resulting vectors in Qdrant, pgvector, or any other storage.
The output is not text or an image, but an embedding: a 1024-dimensional vector that arrives as a standard array of numbers. The input length is limited to a 512-token context, so pre-cut long documents into fragments with a small overlap.
Reasoning is a task for generative models, but gte-large works differently: it does not write answers, but encodes the meaning of text into a vector. Step-by-step reasoning is not noted in our data. To analyze the found fragments, connect a text model from the catalog in tandem.
Function calling and structured output are not noted in our data — these are tools for chat models. However, the response is already strictly formalized: an array of numbers that is placed directly into the database. Requests go through the unified OpenAI-compatible BotHub API.
Accepting anything other than text is not noted in our data: this is a text embedding model. If your project needs multimodality, switch to a suitable model via the same BotHub API — you won't have to rewrite the integration or change keys.
Nothing is said about languages in our data, so we will not promise accuracy in advance. Test it on your own sample: collect control 'query-document' pairs and compare the search quality with other embedders, as switching in BotHub is done via a single API.