This guide explains the differences between the LLMs (Large Language Models) that you can select for Virtual Humans.
The Virtual Human responses are powered by a Large Language Model, which analyzes the learner's input and decides what the Virtual Human says in response. In the character page, we give you the option to decide which model to use, as each one is different and has pros/cons.
The models vary in a few different areas that can affect your virtual human.
Response Time: Affects how long it takes for your Virtual Human to generate a response.
Context Window: The context window is how much information the model can take as an input, which affects how much of the conversation your Virtual Human can remember.
Tone and Style: Some models are trained to be generic and neutral, but can be less "human-like" unless prompted, while others can generate conversational, creative, and artistic responses.
Currently, Virti offers four LLMs, GPT 4, GPT-4o, GPT-4o mini and Claude Sonnet 3.5.
GPT-4: A powerful, versatile model for complex tasks, providing nuanced responses and handling intricate role-play scenarios. It's ideal for deep conversations, though slightly slower due to its advanced capabilities.
GPT-4o(Default): An optimized version of GPT-4, balancing depth and speed. It offers strong performance for moderately complex tasks with faster response times, making it great for more dynamic, real-time interactions.
GPT-4o Mini is a streamlined version of the GPT-4o architecture, optimized for faster interactions. It balances performance and efficiency, making it suitable for quick interactions, simpler role-play scenarios, and straightforward dialogue.
Claude 3.5 Sonnet is a highly capable language model from Anthropic's Claude 3 family. It offers advanced natural language understanding and conversational nuance, ideal for complex role plays and simulations. It can provide enhanced emotional understanding and more realistic responses in longer interactions while maintaining a natural and fluid dialogue.
To change the model, simply navigate to the Character tab, scroll down to AI Settings and choose your model from the drop down. After selecting the model, all new conversations will use this model.
If you have any questions or feedback, contact our support team via support@virti.com and we’ll be happy to help.
