Model catalogue · Google
Gemini Embedding
An embedding model answers nothing - it makes a corpus findable. This tier of the Gemini family turns text into vectors so that a search hits on meaning rather than wording; it is the substructure of file search and of every answer evidenced from your own documents. It is procured via Google Vertex AI, and there the contractual chain is closed: the platform operator's confidentiality agreement under section 203 of the German Criminal Code applies service-bound and covers this model as it does the platform's language models. What remains to be considered is scale: an indexing run sends the entire corpus through the service, and vectors have to be treated like plain text.
Everything section 203 requires is settled, including the provider's confidentiality undertaking, and nothing is stored at the provider, apart from a technical cache of a few minutes for Claude (zero retention). You do not have to clarify anything further.
What it is good for
- Foundation for search and citation matching across your own file corpus
- Callable through the platform's EU endpoint, processing committed to the EU
- Covered by the same service-bound section 203 agreement as the platform's language models
Where it stops
- An embedding run sends the entire indexed corpus through the service - the scope and redaction of the index have to be settled beforehand
- Vectors are not an anonymisation technique: content can be partly reconstructed from them, so they have to be treated like plain text
- Per the vendor documentation the platform's global endpoint gives no residency guarantee - procurement has to be restricted to EU endpoints
- The exception from logging is project-bound - it has to be applied for again per customer project
Status through us
The same model, procured differently, is a different legal position. The status above refers to procurement through us; processing takes place in a third country.
Which route carries it, with which instrument and under which condition, we set out in full to the firm before it commissions us - section 203(4) requires that the chain can be checked, not that it is published. The full matrix, with the sources behind every route, can be requested in the German catalogue after confirming your e-mail address, and we go through it in the conversation.
Is Gemini Embedding the right fit?
We go through your tasks and the route that carries them - and say plainly where it does not.
Arrange a conversation