Independent company · we are not affiliated with, sponsored by, or endorsed by the Swiss AI Initiative (EPFL, ETH Zurich, CSCS), the developers of the Apertus model · Official site: apertus-ai.org
Models & deployment modes
Every model runs behind your dedicated apertus.ai gateway: one API, one permission model, one audit log. For each model you decide whether it runs as an EU cloud model, on dedicated GPUs or on your premises.
One gateway, three modes
01 Modes compared
Your gateway is always dedicated. Underneath, you choose per model how much exclusivity you need, and you can switch at any time without touching your applications.
02 Routing
In the gateway you set rules per team, application or data class. Sensitive requests land on local models automatically, everything else wherever it is cheapest or fastest.
Your gateway
Rules per team & data class
03 Models
We only add models we have reviewed and can run in the EU: leading open-weight models from Qwen, DeepSeek, Mistral, Google, OpenAI and others. New models are added continuously.
Closed-source models in the EU region can be connected on request.
All models & token prices04 Swiss AI Apertus
SwissAI Apertus is the open, multilingual language model of the Swiss AI Initiative. At apertus.ai you use it like any other model: through your dedicated gateway, billed per token and without GPUs of your own.
Swiss AI
Switzerland's first large-scale, open and multilingual language model – ideal for sovereign AI applications.
Prices per 1M tokens, excl. VAT. Usage-based billing.
apertus.ai is an independent company. The Apertus model is developed by the Swiss AI Initiative (EPFL, ETH Zurich, CSCS). We offer it through our gateways and are not affiliated with the initiative.
05 On-premise
When apertus.ai runs on your premises, the MCP gateway connects local models directly to your internal databases, ERP, DMS and file shares. Every access goes through permissions and the audit log, and no request ever leaves your network.
Your network
Your data sources
apertus.ai gateway
AI & MCP, on-premise
FAQ Questions about models
No. Prompts and responses are never used for training. In the cloud they stay in the EU, dedicated models only run on your GPUs, and on-premise nothing leaves your network.
Yes. We run your own or fine-tuned open-weight models as dedicated models in the EU or on your premises and connect them to your gateway.
Through your gateway. Applications always call the same endpoint, and routing decides which model answers where. Switching doesn't require any change to your apps.
That depends on GPU type, term and model. We work out with you at which volume dedicated GPUs beat token prices, and send you a quote.
Open-weight models such as GPT-OSS and Gemma are available directly. Closed-source models in the EU region can be connected to your gateway on request, with the same permissions and audit log.
Yes. SwissAI Apertus is available to all customers through their gateway, at the token price from the price list. The model comes from the Swiss AI Initiative. apertus.ai is independent of it and offers the model through its gateways.
Contact 20 minutes, an honest assessment
Book a short call. We look at your situation and tell you honestly whether and how apertus.ai can help.
What happens in the call