Claude Opus 5.5, Sonnet 5.5 or Haiku 5.5: Which One Should You Use?

A person at a desk thinking while three friendly robots of different sizes, holding a magnifying glass, a notebook and a stopwatch, stand in a row above a row of three checkboxes with the middle one ticked

Written by

in

· Published

Anthropic has released three new Claude models in a few weeks: Opus 5.5, Sonnet 5.5 and Haiku 5.5. If you use Claude, you may have seen them in the model menu and wondered which one to pick. I use Claude for my work and for building my store, but I have not compared these three models side by side, so this is not a review. It is a summary of what Anthropic says in its three announcements, in my own words, and a simple way to choose. I read the announcements with Claude’s help.

The short version

Think of them as three sizes. Opus 5.5 is the most capable and is meant for complex, long jobs. Sonnet 5.5 is the middle option for everyday work. Haiku 5.5 is the small, fast and cheap one for simple, repetitive tasks. If you are not sure, Anthropic’s own descriptions point to Sonnet for most everyday tasks.

The three models side by side

ModelWhat Anthropic says it is forAPI price per million tokens (input / output)Speed claim
Opus 5.5Complex work that needs careful judgment: long coding jobs, research, financial and business analysis.$4 / $20More than 30% faster output than Opus 5.
Sonnet 5.5Well-defined everyday tasks, fixing bugs, and making polished documents, slides and spreadsheets.$2 / $10More than 30% faster than Sonnet 5.
Haiku 5.5High-volume, simple tasks such as summaries, classification, database queries, live customer support and browser use.$0.10 / $0.50 (prompts up to 100,000 tokens)Anthropic’s fastest model at standard speed.

The prices only matter if you use Claude through the API, for example in an app or an automation. If you use Claude with a subscription, you do not pay per token, but you do have usage limits.

What each announcement says

Opus 5.5

  • It is the first model of the 5.5 family. Anthropic says it performs at the level of its previous top model, Fable 5.1, on most work, and that it costs about 40% less to run than Opus 5 on typical workloads.
  • Anthropic also says it writes more clearly than Opus 5, putting the most important information first.
  • With the launch, Anthropic raised the five-hour usage limits on the Pro, Max, Team and seat-based Enterprise plans.
  • Most cybersecurity tasks are re-routed to an older model, Opus 4.8, as a safeguard. Normal software work such as finding and fixing bugs is not affected, according to Anthropic.

Sonnet 5.5

  • Anthropic describes it as a faster, cheaper partner to Opus 5.5, strongest at well-defined everyday tasks and at producing documents, slides and spreadsheets.
  • The price per token is the same as Sonnet 5, but Anthropic says it needs fewer tokens for the same work, so it costs up to 30% less per task.
  • Anthropic is also honest about the limit: it says Opus 5.5 is still clearly better at complex, open-ended work that needs sustained judgment.
  • In the Claude apps, the default effort level is Medium. Lower settings are faster and cheaper, higher settings think longer.

Haiku 5.5

  • Anthropic calls it its cheapest, fastest and most capable small model. It says it costs around 75% less to run than Haiku 4.5 on average.
  • It is the first Haiku with an adjustable effort setting.
  • Anthropic says it is best for narrow tasks like summaries, compaction and sub-tasks inside larger agents, and that Sonnet 5.5 and Opus 5.5 are still the better choice for complex coding.
  • Anthropic also announced a monthly API credit for Max, Ultra and Team subscribers, so they can build their own tools: $100 for Max, $200 for Ultra and up to $500 pooled for Team.

How I would choose

This part is my own rule of thumb, not Anthropic’s, and I have not tested it:

  • Start with Sonnet 5.5 for writing, summaries, spreadsheets, slides and normal questions.
  • Switch to Opus 5.5 when a task is long, has many steps or needs a lot of judgment, and the answer from Sonnet is not good enough.
  • Use Haiku 5.5 only if you build something that runs many small tasks, where speed and price matter more than depth.

I explained how I decide between tools in how I choose which AI to use for each task, and how I use Claude for my own store in how I built my Shopify store with Claude Code.

What to keep in mind

  • The benchmark scores and the price and speed figures in the announcements come from Anthropic itself. I have not measured any of them. Anthropic even says that at these levels, benchmark gaps are a less reliable guide to real differences.
  • Which models you can use depends on your plan and your app. Check the model menu in your own account.
  • Names, prices and limits change. If this article and Anthropic’s pages disagree, trust the pages.

Sources

I read all three pages on November 7, 2026.

About the author
I’m Christian Serrano, and I run AIVERSUSIA. I sell products online and use AI tools in that work every day. I test tools with the same prompts and publish what happened, including mistakes and limits. Claude helps me write articles and run tests, and I review and approve everything before it is published. More about me · Editorial policy