Models | Skills | Agents | Workflows
AutoGen supports custom AI models, we're not limited to using OpenAI's GPT models here.
Notes:
- We can assign custom models to out AutoGen agents, but GPT assistants are not supported yet.
- Changes made to Models do not propagate to Agents or Workflows. Make sure to manually update Agents and Workflows when needed.
- Model name:
gpt-4o - Base URL:
https://api.openai.com/v1/ - Model type:
openai - Description:
OpenAI: Best model, multimodal
More specs:
- Details: https://platform.openai.com/docs/models/gpt-4o
- Context: 128,000 token
- Type: Multimodal
- Model name:
gpt-4-turbo - Base URL:
https://api.openai.com/v1/ - Model type:
openai - Description:
OpenAI: Latest GPT-4 model, vision capabilities
Mode specs:
- Details: https://platform.openai.com/docs/models/gpt-4-turbo-and-gpt-4
- Context: 128,000 token
- Type: Text and Vision
Cheap text generation model; cannot use Skills (or if it can, it's extremely unreliable)
- Model name:
gpt-3.5-turbo - Base URL
https://api.openai.com/v1/ - Model type:
openai - Description:
OpenAI: Simple but fast text model; no Skills
More specs:
- Details: https://platform.openai.com/docs/models/gpt-3-5-turbo
- Context: 16,385 token; max output of 4,096 token
- Type: Text