Beyond Chat is model agnostic: you work with the models of all leading AI providers in one platform. For every task, you pick the right model, or you let the Smart Router decide.
One access point, all leading models
Instead of committing to one provider, in Beyond Chat you choose from the current models by OpenAI, Anthropic, Google, Mistral, DeepSeek, Perplexity and other providers. New model versions are available to you as soon as they are integrated into Beyond Chat, without a separate contract or an additional account.
Which models you actually see is defined by your company in model management.
The Smart Router chooses for you
By default, the Smart Router is active: it analyzes your request and automatically selects the right model. So you never have to think about model names. How this works is explained in the article on the Smart Router.
If you want to decide yourself, simply pick a specific model in the model selection in the chat.
The model info card
Hover over a model in the model selection and you see all its key facts at a glance:
Modalities
Which inputs the model can process: text, documents or images.
Tools
Which capabilities the model has:
Web Search: searches the internet for up-to-date information
Image Gen: generates images
Reasoning: takes extra thinking time for complex tasks
Code Execution: runs code and creates documents
Important: If you want to generate an image or have a document created, choose a model with the matching capability (Image Gen or Code Execution). The Smart Router takes this into account automatically.
Cost (1 to 5)
The relative price level of the model. It is based on the prices in the pricing overview. A low value means: inexpensive for everyday use. A high value means: premium model for demanding tasks.
Intelligence score (1 to 5)
How capable the model is. The value is based on an index of various benchmark tests. The higher the value, the better the model performs on complex tasks such as logical reasoning and problem solving.
Speed (tokens per second)
How fast the model responds, measured in output tokens per second. Reasoning models are slower because they use extra thinking time before answering.
Hosting location
Where the model is hosted, for example the EU. Relevant for GDPR compliance and your company's compliance requirements.
Knowledge cutoff
The date up to which the model was trained. The model knows nothing about events after that date.
Tip: If you need up-to-date information, enable web search. The model then researches live on the internet, regardless of its knowledge cutoff.
Context window
How much text the model can process at once, measured in tokens (a token corresponds to roughly 1.3 words). A model with a 200K context window can process entire books or extensive document collections in a single conversation.
Keep in mind: the context window includes not just your current message, but the entire chat history and all uploaded documents. The fuller the window, the more the answer quality suffers.
Tip: Open a new chat for every new topic. This keeps the context window lean and the answers high quality.
Related articles
Questions? Message us right here in the chat.