Heads up: some links below are referral links. They cost you nothing extra and never change what we recommend. How this site makes money.
The short answer
Use Sonnet 5 by default. It is the daily driver — strong reasoning, fast enough for real-time work, and capable enough that most tasks never outgrow it.
Switch to Opus 5 for genuinely hard reasoning and long agentic tasks. It is stronger, and it consumes your usage allowance considerably faster, so it is worth saving for work that needs it.
| Haiku | Sonnet 5 | Opus 5 | Fable 5 | |
|---|---|---|---|---|
| Best for | Quick, simple tasks | Everyday work | Hard reasoning, agents | Specialised work |
| Speed | Fastest | Fast | Slower | Slower |
| Usage consumed | Lightest | Moderate | Heavy | Heaviest |
| Use as default? | For simple queries | Yes | No | No |
Why Sonnet is the default
Sonnet 5 is designed as the model you use for nearly everything: coding, writing, analysis, research and complex problem-solving. It is responsive enough for back-and-forth work and capable enough that most problems do not exceed it.
Its performance is close to the previous generation's top-tier model while costing considerably less to run — which, on a subscription, translates directly into how long your allowance lasts.
For most people, most of the time, the difference between a Sonnet answer and an Opus answer is not visible. The difference in how fast you hit your limit very much is.
When Opus earns its cost
Opus 5 is stronger at deep reasoning, long-horizon agentic tasks, and problems that benefit from more thinking time. Reach for it when:
- A task involves many interdependent steps where an early mistake compounds.
- You are running a long agentic session across a large codebase.
- Sonnet has already tried and produced something subtly wrong.
- The reasoning genuinely is the hard part — not the writing, the thinking.
That last one is the practical test. If the task is hard because it requires careful multi-step reasoning, Opus. If it is hard because it is long or tedious, Sonnet handles it fine.
The usage cost nobody mentions
This is the part that matters on a subscription. The models consume your allowance at very different rates — Haiku lightest, Sonnet moderate, Opus heavy, Fable heaviest.
Setting Opus as your permanent default is the single most common reason people hit Claude's usage limits and conclude they need to upgrade. Very often they do not; they need to switch models.
Try this before upgrading: use Sonnet 5 for a week and only switch to Opus when a task actually stumps it. Many people find their limit problems disappear entirely.
A simple rule
Default to Sonnet 5. Escalate to Opus 5 when Sonnet visibly struggles.
That is genuinely the whole heuristic. It costs you one retry on the occasional hard task and saves an enormous amount of allowance on the many easy ones.
Haiku is worth remembering for quick lookups and simple summaries, where it returns answers almost instantly and barely touches your usage.
Common questions
Is Opus always better than Sonnet?
Which model does Claude use by default?
Does using Opus use up my limit faster?
What is Fable for?
Both models come with any paid plan
You don't pay separately for Opus or Sonnet — switching between them is just a dropdown, and using it well makes your allowance last much longer.