Choose the model around the work and the data
Quality, location, licensing and operating costs matter more than a single provider preference.
Editorial analysis and illustrative technical scenarios. They do not define the features or integrations included in a UTOMAI subscription. See Private AI On-prem pre-launch information.
AI-generated editorial illustration. Not a product photograph or a customer case study.
A company’s AI needs rarely fit one model for every task. Drafting, document consultation and specialist assistance place different demands on quality, response time and data handling. Begin with those requirements, then assess the model options available in the platform configuration.
Open weights are the intended foundation
The planned UTOMAI catalogue draws on open-weight families such as Mistral, Qwen, Meta Llama and DeepSeek. The exact models, revisions and licences must be confirmed for the selected service. A family name does not promise every size or every capability on every appliance.
The origin of a model and the location of inference are separate questions. Downloadable weights may be run in a chosen environment when the licence and technical configuration allow it. A hosted endpoint needs its own location and data-flow review, regardless of the model’s name.
External frontier models remain a choice
Where a requirement justifies it, an external frontier model from a European or non-European provider can be assessed on request. Agree the destination, terms and additional costs before sending company material. It should be an explicit option rather than an invisible fallback.
Use the same representative questions and review criteria when comparing models. Include difficult documents and requests that should be declined. The aim is a model choice the team can explain and reassess while continuing to work in its private AI platform.
