BespokeSovereign inference

Sovereign inference for your applications.

From our servers in Switzerland, we serve recent open-weight AI models, made available by their developers, behind an API compatible with OpenAI’s: your applications change their address, not their code. For any company building its own AI tools. If your applications need Swiss law, add the legal API.

Bespoke pricing, depending on the models and volumes your project needs.

Request a quote

What you get

Leading open-weight models, at your service.

  1. Recent models

    We follow open-weight releases and put the models we select into service within days of their release, after evaluating them on our own benchmarks.

  2. Up to a million tokens of context

    Depending on the model, enough to submit an entire file in a single request. Verified: in texts of 256,000 to 891,000 tokens, the 24 pieces of hidden information were all retrieved.

  3. Text, images and voice

    Generation and reasoning, image analysis and speech transcription, through the same API. Embeddings and reranking for your own search engines, on request.

  4. Your own models

    A model you have trained or chosen: we explore with you how to deploy it on capacity dedicated to you, on our infrastructure or on your own server.

Sovereignty

Your data does not leave Switzerland.

Your requests are processed on our servers, in Switzerland, without any external AI provider. Each application gets its own access key, and its usage is metered separately.

Depending on the hardware chosen, the same models also run on your own server, which we install and maintain.

Bespoke pricing: together we define the models, the volumes and the contract suited to your project, then send you a quote.

Request a quote