Skip to content

AI glossary

Best Claude model: which one to use for what

Claude comes in four sizes. As of November 2026 they are Haiku 5.5, Sonnet 5.5, Opus 5.5 and Fable 5.2, the newest and most capable.

By Published

Dani runs four apps with AI and has trained designers at PepsiCo.

The short answer
  • There isn’t one best model. There’s a best one for each job.
  • Most people, most days: Opus 5.5.
  • Biggest projects: Fable 5.2. Quick everyday work: Sonnet 5.5. Simple jobs in bulk: Haiku 5.5.

The four Claude models, smallest to largest

Haiku 5.5

The quick one

  • Fastest and cheapest
  • Simple, repeated jobs
  • Sorting, tagging, short replies

Sonnet 5.5

The everyday one

  • Fast
  • Drafts, summaries, rewording
  • Questions about a file

Opus 5.5

The daily driver

  • Where Anthropic says to start
  • Coding and analysis
  • Work that needs judgment

Fable 5.2

The heavy lifter

  • Newest and most capable
  • Projects that run for hours
  • Directs and checks smaller models

What each one costs

Haiku 5.5from $0.10
Sonnet 5.5$2
Opus 5.5$4
Fable 5.2$8

Price per million tokens going in, through Anthropic’s API. Output costs five times as much on every model. On a Claude plan you don’t pay per token.

Which one for which job?

The jobUse
Tagging 2,000 customer messagesHaiku 5.5
Drafting an email or a postSonnet 5.5
Summarizing a long documentSonnet 5.5
Coding, analysis, anything needing judgmentOpus 5.5
A plan with many parts, or a final reviewOpus 5.5 or Fable 5.2
A project that runs for hours on its ownFable 5.2
From my own business

Since July I’ve run my business on one rule: the big model is the boss, the smaller one is the worker, and the boss checks the work.

Ordinary messages go to the everyday model. If I ask it to think something through, it switches up.

A cheaper model can do the work. It never gets the last word on what reaches me.

Three mistakes I see people make

  1. Using the biggest model for everything.You pay more, wait longer and hit limits sooner.
  2. Letting a small model check its own work.The check should come from a stronger model, or from a plain rule.
  3. Switching models before fixing the request.A thin answer usually means a thin request.

How to choose, step by step

Start in the middleSonnet 5.5 or Opus 5.5.
Thin answer?Add the missing background first.
Still thin?Move up one size for that job.
Want help setting this up in your business?

Talk it through with me. Tell me what your work looks like and we’ll find where AI helps first.

Talk it through with Danielle

Quick answers

What is the best Claude model?

For most work, Opus 5.5. Anthropic recommends starting there.

For the biggest and longest projects, Fable 5.2 is the most capable.

What is the latest Claude model?

Claude Fable 5.2, released on 9 November 2026. Opus 5.5, Sonnet 5.5 and Haiku 5.5 are the current smaller models.

Which Claude model is the cheapest?

Haiku 5.5, from $0.10 per million tokens in and $0.50 out. It is also the fastest.

Which Claude model is best for writing?

Sonnet 5.5 for most writing. Move up to Opus 5.5 for a long piece with a lot to hold together, or for the final review.

Which Claude model is best for coding?

Opus 5.5 for most coding. Fable 5.2 for long projects with many parts, where it scores highest on Anthropic’s hardest coding test.

Do I need the most powerful model?

Usually not. A thin answer more often means a thin request than a small model. Fix the request first, then move up.

Next: Fable 5.2 vs Opus 5.5 in detail, Opus vs Sonnet, and what an LLM is.

AI Foundations starts here: give AI a clear picture of your business in a form you can reuse, so you stop explaining from zero.

Background: Anthropic: Introducing Claude Fable 5.2 · Claude docs: model comparison and prices. Figures checked 10 November 2026.