I'm using text-embeddings-inference by huggingface, with model id = yco/bilingual-embedding-base. They provide 2 endpoints for embedding, among which one is supposed to be on par with openai