OpenAI's text-embedding-3-small costs only about 1 dollar / 50 mio tokens.

https://huggingface.co/Qwen/Qwen3-Embedding-0.6B is open-source

Also is Nomic, which is only 250 MB: https://ollama.com/library/nomic-embed-text/tags

Nomic is quite fast an can be used in Python via Ollama:

ollama.embeddings(model='nomic-embed-text', prompt='Hello world')

Benchmarks for embeddings are often available at:
https://huggingface.co/spaces/mteb/leaderboard

Languages differences

Many embedding models are trained on multiple languages.
Among them is: https://huggingface.co/intfloat/multilingual-e5-base also getting good results for german text.

Depending on the target language, consider a different embedding model.

Professional context / tone

When dealing with professional texts, e. g. technical documentation, selection of the embedding models matters even more.

• T-Systems-onsite/cross-en-de-roberta-sentence-transformer