How models differ

GPT models (OpenAI) are literal. They do exactly what you wrote, and little more. If something matters, say it: every requirement, every file, what to do when done. A short request gets a short, narrow answer.

Claude models (Anthropic) fill gaps. They work out what you probably meant and often do more than you asked. That saves typing, but say clearly what is out of scope, or you may get extras you did not want.

So the same prompt can give very different results. When one family misreads you, do not just repeat yourself: add the missing detail for a literal model, or the missing limits for a gap-filler.

• For GPT: number the requirements, name every file, and say "do all of the above, then report".
• For Claude: say what success looks like and what to leave alone; it will find the steps.

Start with the default. Move to a bigger model when the work is hard or the answers are shallow, and to a smaller one for quick, simple jobs.

Some models let you choose how hard they think. More effort is slower and uses more, but helps with hard reasoning, tricky bugs and long plans. Quick edits do not need it.

You can change model at any point in a chat, even to another family. The new model is handed the conversation so far and carries on.