Anthropic shipped three new models in three weeks. A plain guide to the differences, and how to choose between them for real business tasks.
Between September 22 and October 7, Anthropic released Claude Opus 5.5, Claude Sonnet 5.5 and Claude Haiku 5.5. If you use Claude in a browser, you may simply have noticed it got faster. If you are building something with it, the choice of model now matters a lot for cost.
Here is how we think about the three.
Opus 5.5: the careful senior
Released September 22. Anthropic describes it as performing at the level of its larger Fable 5.1 model on most work while costing about 40% less to run than Opus 5. It is priced at $4 per million input tokens and $20 per million output tokens, with a one-million-token context window.
Use it for work where being wrong is expensive: planning a software change across a large project, reviewing a contract summary before a person reads it, or working through a long, messy set of documents.
Sonnet 5.5: the everyday workhorse
Released September 28. Anthropic says it is faster than Sonnet 5 and typically needs fewer tokens to do the same job. On October 7 the price of reading from its prompt cache was halved, which makes repeated tasks with the same instructions cheaper still.
Use it for most business tasks: drafting, answering questions from your own documents, writing and fixing code, building spreadsheets and slides.
Haiku 5.5: fast and very cheap
Released October 7. Anthropic calls it its fastest model and says it costs about 75% less to run than Haiku 4.5. It also has an effort setting, so a builder can trade a little speed for more thinking on harder requests.
Use it for high-volume, simple jobs: sorting incoming messages, tagging support tickets, pulling a date and a name from a form, or acting as a quick helper inside a bigger workflow.
A simple way to choose
- If a person will read every output anyway and the task is routine, start with Haiku.
- If the output needs to be good on the first try, use Sonnet.
- If the task is long, high-stakes or spans a lot of material, use Opus.
- Mix them. A common pattern is Haiku sorting the work and Sonnet or Opus handling the few items that need real thought.
The part that didn’t change
Better models make good processes faster. They don’t fix a process nobody has thought through. Before choosing a model, write down the task in one sentence, what a good result looks like and who checks it. That page will save you more money than any price cut.
Sources
From the DYOL blog. Photo by Laura Ockel on Unsplash.
Keep reading
All posts
Claude now makes docs, decks and designs in any chat
September brought Cowork to every conversation, memory across chat and Cowork, and a Salesforce connection in beta. What that means for everyday office work.

Claude can now use a computer. Here is why that matters
Anthropic’s computer use tool left beta in August, alongside a new browser tool. What it means for businesses stuck with software that won’t connect to anything.

What’s new in Claude Code: the late-summer releases that matter
Sessions that talk to each other, design in the terminal, background computer use and a way to test plugins. Here is what changed and why it matters.
A free 20-minute conversation, then an honest assessment. You’ll know what it costs and how long it takes, in writing, at a fixed price.
Or email lloyd.delacruz@outlook.com. We reply within 24 hours.
