Choosing a model

Claude comes in more than one model. Anthropic releases new ones regularly and retires old ones, so any list printed in a lesson goes stale. This lesson teaches the way to choose, and sends you to the official model overview at docs.claude.com for the current names, limits, and prices.

The tiers

Anthropic has used a consistent pattern of tiers, and as of this writing the current lineup still follows it:

• Haiku: the smallest and fastest tier, aimed at quick, high-volume, lower-cost work.
• Sonnet: a balance of speed and capability for everyday work.
• Opus: a more capable tier for harder reasoning and longer, more complex tasks.

There are also newer top-end models above these at times. Names such as "Opus 5.5" or "Sonnet 5.5" identify a generation within a tier. Newer generations are usually better at the same tier, so prefer the latest unless you have a reason not to.

In the chat apps, you may see a model picker. Which models appear depends on your plan. In the API and Claude Code you choose by an exact model ID, and the docs list the current IDs. Copy them from the docs rather than from memory or a blog post, since IDs are exact strings and a typo means an error.

A simple decision rule

Ask three questions.

1. How hard is the task? Rewriting a sentence is easy. Untangling a legal argument, debugging a subtle program, or planning a multi-step project is hard. Hard tasks benefit from the more capable tier.
2. How fast do you need it? Smaller models respond faster. For quick back-and-forth or many small jobs, speed matters.
3. What does a mistake cost? If an error is cheap and you will review it anyway, a smaller model is fine. If an error is expensive, use the stronger model and still review.

Thinking and effort settings

Some models can spend extra effort reasoning before they answer, and some apps expose that as an "extended thinking" toggle or an effort setting. More thinking usually means slower and costlier answers, and it helps most on math, logic, code, and multi-step planning. It helps little on casual writing. The exact controls change between versions, so read the current docs.

Do not chase the biggest model by default

A common beginner habit is to always pick the top option. Try the opposite for routine work: start with a smaller model, and move up only when the answer is not good enough. You learn quickly where the line is for your tasks, and you save usage.

Test on your own work

Benchmarks and rankings tell you about averages. Your tasks are specific. The best comparison is to run your own real prompt on two models and read both answers.

Try it

Pick one real task of medium difficulty, such as "turn these meeting notes into action items with owners." If you have a model picker, run the same prompt on two different models. Score each answer from one to five on accuracy, tone, and how much editing it needs. Write a one-line note: which model you will use for this task and why. If you only have one model available, run it once with a short prompt and once with extended thinking on, if offered, and compare.