These are like the Toyota Camrys of AI models - pretty good at a broad range of things.
All of the popular AI chatbot tools like ChatGPT, Gemini, Claude, etc. have general-purpose models. They aren’t always great at identifying them, though.
Here are common tasks where you would use a general-purpose model -
- Simple question and answer
- I.e., What is the capital of Australia?
- Single instruction tasks
- Summarization
- Language translation
- Conversation
- Content generation
Pros of general-purpose models
- Availability - these are the easiest models to find. Just about every AI chatbot that you find will have one.
- Versatile - they can handle a diverse range of tasks.
- Accurate - for common tasks.
- Multimodal (sometimes) - fancy word meaning that they can handle more than just text (i.e., images, audio, or video).
- Cost - these models can be less expensive to use than reasoning or deep research models.
Cons of general-purpose models
- Hallucinations - the classic AI issue of making up false or nonsensical information.
- Not good with logic - these models have issues with math, logic, and reasoning tasks.
- Not good with complex tasks - complex in this context meaning tasks with a lot of steps.
- Bias - output will be biased based off of the training data used and the people that tuned the models.
In general (ha), you want to start with these types of models and switch to a different type when you consistently get unusable output (unless your prompts are bad, but that’s a different post…)
Comments