> ## Knowledge Base Index > Fetch the complete knowledge base index at: https://help.chrisp.chat/sitemap.xml > Use this file to discover available pages before exploring further. > Pure-Markdown content can be obtained by appending a '.md' suffix to the content URLs listed in the sitemap (without the trailing slash). # Which AI models are supported by Hugo __This article explains which AI model families Hugo can use and how to choose the right one for your workspace.__ Hugo supports several model providers so teams can balance answer quality, speed, cost, privacy, and compliance requirements. Model availability can evolve, so always confirm the current list from ****AI Agent → Agent → Settings**** before making procurement, compliance, or internal policy commitments. --- # Available model families #### OpenAI GPT models OpenAI models are strong general-purpose options for support conversations, summaries, knowledge-based answers, and complex reasoning. They are often a good default when you want high-quality answers across a wide range of topics. #### Anthropic Claude models Claude models are useful for nuanced conversations, longer context, and cases where answer style and careful reasoning matter. Claude Haiku-style models are generally lighter and faster, while Sonnet-style models are better suited for more complex exchanges. #### Mistral models Mistral models are a strong option for multilingual support and teams that prefer European model providers. They are especially relevant when your internal policies require careful consideration of data processing location and provider choice. #### Google Gemini models Google Gemini is available for teams that want to test or use Google's model family inside Hugo, depending on availability in their workspace. #### Crisp Internal Model The Crisp Internal Model is hosted by Crisp in France. It is designed for teams with stricter privacy or compliance requirements, or for use cases where keeping processing within Crisp-hosted infrastructure is important. --- # How to choose a model There is no single best model for every workspace. The right choice depends on what Hugo needs to handle and how your team balances performance, cost, speed, and compliance. __A good model evaluation usually checks:__ * ****Answer quality**** → how accurately Hugo answers your most common questions * ****Reasoning quality**** → how well Hugo handles multi-step or ambiguous requests * ****Speed**** → how quickly customers receive an answer * ****Credit usage**** → how the model affects your monthly Hugo usage * ****Compliance needs**** → whether your business has internal rules about providers or data location * ****Language coverage**** → how well the model handles your users' languages Use ****AI Agent → Evaluate → Playground**** to compare models with the same questions. Test real customer messages, not only ideal examples, and review the Guardrails to understand how Hugo searched and answered. --- # Privacy and EU compliance If privacy or EU compliance is a priority, review model choice carefully. The ****Crisp Internal Model**** and ****Mistral**** are usually the first options to evaluate for EU-focused requirements, but the right choice depends on your internal policies and the type of data Hugo handles. In addition to the main model you select, Hugo can rely on internal Crisp models for supporting tasks such as safety checks, spam handling, escalation decisions, and other guardrails. These supporting models help Hugo keep answers grounded and decide when a human handoff is safer. | Model availability can change over time. Use the model selector in ****AI Agent → Agent → Settings**** as the source of truth for your workspace.