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.

EmbeddingsGigaR is an embedding model from GigaChat that converts text into a numerical vector. It is suitable for semantic search, RAG on knowledge bases, document clustering and deduplication, ranking, and classification of requests. You store the vectors in Qdrant, pgvector, or another storage and compare them using cosine similarity.

The output is not text, but a fixed-dimension vector — an array of numbers; see the API response for the exact dimension value. The model context is 4096 tokens, which is also the maximum output, so pre-chunk long documents into appropriately sized fragments.

Reasoning before answering is not indicated in our data, and it is not required for embeddings: the model does not write an answer but encodes the meaning of the text into a vector. Leave reasoning logic to the generative model to which you pass the fragments found via EmbeddingsGigaR.

Function calling and structured output are not indicated in our data — these are tools for generative models. However, the API response itself comes structured: an array of numbers that you write directly to a vector database. Through the unified OpenAI-compatible BotHub API, the request looks familiar.

Data input other than text is not indicated in our data, so expect text input. If you need to vectorize the content of audio or images, first get a text description using another model in the same BotHub window, and then pass it to EmbeddingsGigaR.

The model was created by the Russian vendor GigaChat and is focused on working with Russian-language corpora, so vectorizing such texts is its primary scenario. There are no specific accuracy ratings in our data, so check the quality for your specific domain on a small test sample of queries.