Introduction to the LLM Model Guide

Each Mate runs on a large language model (LLM). The model shapes the quality of the replies, their speed and the credits that each reply consumes. This guide helps you pick the right model for each Mate. It is for everyone who configures Mates, and for the organization Owners and Admins who decide which models the organization offers.

Why the model matters

• Quality: models differ by task. One model writes well, another one reasons better on data or code.
• Speed: a smaller model replies faster. A larger model or a higher thinking effort takes more time but can solve harder tasks.
• Credits: a larger model consumes more credits per reply than a smaller one.

Match the model to the task. A light model is often enough for short answers and simple questions. Keep the larger models for complex work.

What you find in this guide

• Current lineup: the models that your organization can use today, the Allmates models, the "Latest" models that update themselves, the deprecated models, and links to public leaderboards to compare models.
• Managing your organization's models: for Owners and Admins. Add models to the organization, set the default model and remove the models nobody uses.
• Model Fallback: what happens when a model does not reply, and how to choose the backup model.

To set the model of one Mate, open the Mate, then its Parameters tab. The Model & generation card holds the Model list and, depending on the model, the Temperature slider and the Thinking effort. See "Configuring Mate Capabilities".

Related articles

• Current lineup
• Managing your organization's models
• Model Fallback
• Configuring Mate Capabilities
• Understanding credits