How to pick AI modelsfrom @claudeai

Transcript

Here's how to pick the right model in cloud code. Sonic is the everyday default. It's best, and it's a great generalist for jobs like checking chunks of code before you ship it. But when the work begins to get a little more complex, I start using Opus. Great for jobs like trying to solve tricky bugs or even poking holes in system designs before I even start building. And that's because Opus is really good at deep reasoning, which is why most of my planning starts here. But next up, there's Fable, which is designed to tackle your hardest work. The difference between Fable and Opus is that Fable needs way less step by step guidance. Give it a big task like add a user settings page to my app, which sounds simple, but every toggle has to map to a real database field and user permissions. Fable understands that complexity and will put a plan in place, work through that plan, and deliver you a finished feature. But picking the model is only half of it. If Cloud Code's responses aren't landing the way you hope they would, try tweaking the effort level. This controls how thorough Cloud Code should be and how long it should work on your request. So slide it towards low when you need a fast answer, and up the effort when you want Cloud Code to think longer, run tests, and double check its work. So what model do you find yourself using most?