Model catalogue · Alibaba Cloud (Qwen)

Qwen Embedding

The embedding model of the Qwen family, designed for long inputs and more than a hundred languages: it classifies longer file passages in one piece instead of cutting them up first, which preserves the context of a passage and makes document search more accurate. The gain in accuracy costs compute time and space. The contractual chain is closed in dedicated operation on reserved hardware in a German data centre: the weights are open, there is no model vendor anything is transmitted to, and the data centre operator is bound via the chain obligation under section 203(4) of the German Criminal Code. Via IONOS the product-specific confirmation of the section 203 clause is so far missing.

Available through us

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

  • 32,768 tokens of input - longer file passages are classified in one piece
  • 8 billion parameters in BF16, around 19 GB - runs alongside a language model on one card
  • Available through German data centres and on the dedicated instance, Apache 2.0

Where it stops

  • Considerably larger than the small embedding models - the gain in accuracy costs compute time and space on the card
  • Embedding answers no question, it arranges the corpus for the language model
  • The product-specific confirmation of the section 203 clause for the AI Model Hub is still outstanding

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 Germany.

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 Qwen 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