Models

All leading AI models.
One secure workspace.

Use models from OpenAI, Anthropic, Mistral, Google and more in the same chats, assistants and agents – or let Auto (EU) pick the best fit for every task. Processed in the EU, never used for training, billed without margin.

MODEL CATALOGUE

Choose a model –
or let Auto (EU) choose.

Filter by provider or by what the model should do. Admins decide which models are enabled in your organisation.

Auto (EU)

Default

Not sure which model fits? Auto (EU) routes every request to a model that suits the task – balancing quality, speed and cost – and stays with models hosted in the EU.

See it in Chat

For example

  • Summarise a 120-page contractClaude Opus
  • Translate a short customer emailAuto (EU)
  • Analyse a sales spreadsheetAuto (EU)
  • Draft a product FAQ in FrenchAuto (EU)

Provider

Capability

11 models

  • AnthropicEU – Hosted and processed in the EU

    Claude Opus

    Anthropic’s most capable family, for deep analysis and careful writing.

    • Text
    • Vision
    • Reasoning
    • Code

    Recommended for: Long documents, complex analysis

  • AnthropicEU – Hosted and processed in the EU

    Claude Sonnet

    A balance of speed and quality for everyday work and coding.

    • Text
    • Vision
    • Reasoning
    • Code

    Recommended for: Drafting, research, code

  • AnthropicEU – Hosted and processed in the EU

    Claude Haiku

    Fast and lightweight for simple tasks at volume.

    • Text
    • Vision

    Recommended for: Classifying, extracting, short replies

  • OpenAIEU – Hosted and processed in the EU

    GPT-5

    OpenAI’s flagship for broad knowledge, analysis and code.

    • Text
    • Vision
    • Reasoning
    • Code

    Recommended for: Data analysis, structured answers

  • OpenAIEU – Hosted and processed in the EU

    GPT Image

    Creates and edits images from a prompt, right in the chat.

    • Image generation

    Recommended for: Visuals for posts, slides and drafts

  • Mistral AIEU – Hosted and processed in the EU

    Mistral Large

    Made in Europe

    Mistral AI’s top model – built in Europe, strong in European languages.

    • Text
    • Reasoning
    • Code

    Recommended for: Multilingual writing, reasoning

  • Mistral AIEU – Hosted and processed in the EU

    Mistral Medium

    Made in Europe

    Solid quality at a lower cost, with image understanding.

    • Text
    • Vision

    Recommended for: Everyday assistants, documents

  • Mistral AIEU – Hosted and processed in the EU

    Mistral Small

    Made in Europe

    Small and quick for routine tasks at scale.

    • Text

    Recommended for: Translations, short summaries

  • GoogleEU – Hosted and processed in the EU

    Gemini Pro

    Google’s multimodal model for long inputs, images and reasoning.

    • Text
    • Vision
    • Reasoning
    • Code

    Recommended for: Large files, mixed media

  • GoogleEU – Hosted and processed in the EU

    Gemini Flash

    A fast multimodal model for quick answers.

    • Text
    • Vision

    Recommended for: Quick questions, image descriptions

  • Local AIOn-site – Runs on your own hardware

    Open-weight models

    On your hardware

    Open models selected for your use cases, running on a Mac mini or Mac Studio in your building.

    • Text
    • Code

    Recommended for: Data that must not leave the premises

    Explore Local AI

Admins decide which models are enabled in your organisation; availability can differ by plan and deployment. Model families are updated as providers release new versions.

One request, the right model. Auto (EU) weighs quality, speed and cost for every task – so teams get good answers without knowing the model landscape, and simple jobs don’t run on the most expensive model.

  • FIG 0.1

    Best fit per task

    Complex analysis goes to a strong reasoning model, quick replies to a fast one.

  • FIG 0.2

    0 % margin on tokens

    Model usage is passed through at cost – and Control Tower shows every euro of AI spend.

  • FIG 0.3

    You stay in charge

    Admins decide which models are enabled; anyone can still pick a model by hand.

SOVEREIGNTY

Every model, the same rules.

Processed in the EU

Application, storage and processing run in the EU, operated by a European provider.

No training on your data

We use enterprise APIs – your prompts and files are never used to train models, whichever model you pick.

Governed in one place

Permissions, usage and cost for every model in Control Tower – no separate accounts per provider.

YOUR OWN MODELS

Bring your own models.

For data that must stay on-site, for existing provider contracts and for dedicated capacity.

Local AI

Open-weight models on a Mac mini or Mac Studio in your building, connected to 48°.

Your provider contracts

Already work with a model provider? Talk to us about connecting your own keys and endpoints.

48° Cloud

COMING SOON

Dedicated model endpoints in a sovereign EU cloud setup, for large organisations and regulated workloads.

FAQ

Questions about models

NEXT STEP

Give every team the right model.

In a short demo we show you Auto (EU), the model line-up and the admin controls.