The Model

Your agent is an LLM model (for e.g; chatgpt ), you provide a text, and based on its context, it gives you an outcome, based on patterns it was trained on. Choosing the model is the most important choice you have to make, as it will decide how your agent behaves, thinks, responds and takes actions. You can also chose multiple models for different part of your agent. Here are few points which may help you :

1. GPT model - gpt models by openai are good at reasoning, they are extremely good at that, but that’s all. They are not good with anything tech, you may have noticed this while using chatgpt vs claude.

2. Anthropic - anthropic models are best when you need to perform technical actions, there technical thinking ability is good

3. LLAMA - it’s an open source model, which powers majority of models out there --- gpt and claude are good at reasoning but because they are censored models, you can’t customize them a lot which can result in bot’sh behavior. For cheap inference, do check Inference by Kuzco.

4. BYOM - you can choose to bring your own model and train on cluster providers like ionet which provides gpu’s for training.