> ## Documentation Index > Fetch the complete documentation index at: https://docs.orbitagents.xyz/llms.txt > Use this file to discover all available pages before exploring further. # Which Claude to use > The fast one, the smart one, the everyday one — and a simple rule for picking. Claude comes in a few versions. Think of them like hiring: a quick junior, a solid all-rounder, and a senior expert. You don't need to memorize anything — there's a simple rule at the bottom. ## The lineup, in plain terms * **The everyday one (Sonnet)** — smart, fast, and the right choice for almost everything. This is your default. * **The expert (Opus)** — the deepest thinker. Slower and pricier, but worth it for hard, important, or complicated work. * **The quick junior (Haiku)** — fast and cheap. Perfect for simple, repetitive jobs done in bulk. * **The newest (Fable)** — the latest top-tier model when you want the most capable Claude available. ## The simple rule > **Use the everyday one by default. Switch to the expert when something is hard or really matters. Use the quick junior for simple, repeated tasks.** That's genuinely all you need. ## Where this saves you money It matters most for **scheduled agents** (the ones that run on a timer). An agent running every hour on the expert model gets expensive fast — and usually it doesn't need to. For routine jobs like a daily summary, the quick junior does the job for a fraction of the cost. A good habit: * Runs often + simple → **quick junior** * Runs daily + normal → **everyday one** * Runs occasionally + complex → **expert** ## You can change your mind anytime Because every run is saved to your Orbit Vault, you can look at what an agent produced and decide it needs a smarter (or cheaper) model next time. No commitment — just dial it up or down. Why Claude asks permission — and when to let it run free.