AI

What Is Claude? The Claude Models and How to Use Them

Logo Claude dạng đèn neon cam phát sáng trong không gian tối, chữ Claude trắng bên dưới
In this article

    Fable 5.2 costs 80 times as much as Haiku 5.5. That number is right on Anthropic’s price list: $8 versus $0.10 per million input tokens. Pick the wrong model and you can pay ten times more for a job a cheaper model would do just as well. Hand a hard job to a model that is too small, and you get a half-baked answer.

    This article helps you avoid both. I explain what Claude is, how the four models differ, and which model to give each kind of job.

    What is Claude?

    Claude is an AI assistant built by Anthropic. You can use it to write emails, explain code, summarize documents, analyze images, or work with data. It handles both text and images, and it fits a wide range of jobs — from simple everyday requests to complex technical tasks.

    One detail worth noting is that Claude can take in a large amount of information in a single session, thanks to a context window of up to 1 million tokens on some models. Anthropic estimates that is roughly 555,000 words. That means you can feed it a large set of documents or a project’s source code and have it analyze the whole thing, instead of splitting everything into small pieces.

    Anthropic built Claude as a family of models rather than a single version. Each model has its own strengths, response speed, and cost, made for different needs. Some are a good fit for quick, simple tasks, while the stronger ones are built for work that needs deeper reasoning and heavier processing. Which model you pick affects both your experience and what you pay.

    Four models, four roles

    The table below was verified in November 2026.

    ModelBest forSpeedAPI price (USD / 1 million tokens, in / out)
    Fable 5.2Hard problems that need deep reasoning and run on their own for hoursModerate$8 / $40
    Opus 5.5Writing code and long-running knowledge workModerate$4 / $20
    Sonnet 5.5Everyday work that needs to be both fast and goodFast$2 / $10
    Haiku 5.5Simple jobs repeated thousands of times: classifying, extractingFastest$0.10 / $0.50*

    *For prompts over 100,000 tokens, Haiku is priced at $0.50 in / $2.50 out.

    The easiest way to remember them is to picture a team:

    • Haiku is the person who knocks out repetitive jobs. Hand it 5,000 customer messages to label as “praise / complaint / price question” and it is done before your coffee is.
    • Sonnet is the colleague you call first every morning. Fast enough to feel almost instant, good enough for most writing and coding.
    • Opus is the more experienced one. You turn to it when the job spans many files or Sonnet is not getting it right.
    • Fable is the specialist for the really hard problems, able to run for hours with nobody sitting next to it. The price is that it burns the most tokens.

    Besides these four there is also Mythos 5.2, but it is only available to organisations verified by Anthropic, so ordinary users will not see it on claude.ai. Older models such as Fable 5.1 and Opus 4.8 still exist, but for new work start with the four in the table.

    Which model should you use?

    Anthropic suggests starting with Opus 5.5. In my own experience it works better from the bottom up: start with Sonnet and only move up when it is not doing well. Sonnet is faster and cheaper, and for most jobs you will not see a difference.

    The rule is simple: the clearer and more repetitive the job, the lower the tier you pick. Only the vaguer, harder jobs call for a higher one.

    • Drafting emails, short documents, or asking for an explanation of a concept: Sonnet 5.5.
    • Fixing a bug or building a feature that touches many files: try Sonnet first. If it goes in circles, move to Opus 5.5. Only go up to Fable 5.2 when Opus still cannot handle it.
    • Labelling or extracting information from thousands of records: Haiku 5.5. The answers are easy to check, so speed and price matter more than depth.

    Where to use Claude

    There are three doors, depending on what you are doing:

    claude.ai is the place to chat, ask questions and draft text. Open your browser, sign in and type into the chat box. The model you are using shows right on the chat box, and you click it to change.

    Claude Code is a coding tool that works directly on the files in your project instead of only answering in a chat box.

    The API is for developers who want to put Claude inside their own application, for example a customer-support chatbot or a tool that summarises documents automatically. API stands for application programming interface. The prices in the table above are the prices for this door.

    If you are trying Claude for the first time, just open claude.ai. You do not need the other two yet.

    Frequently asked questions

    Is Claude free?

    Yes. You can chat for free on claude.ai with usage limits. Paid plans raise those limits.

    Is the newest model always the best?

    No. Fable 5.2 is the strongest, but for everyday work Sonnet answers faster and costs far less. A stronger model is only worth paying for when the job is hard enough.

    Can I still use older models?

    Yes, but Anthropic lists them as legacy. For new work, start with the four current models.

    Conclusion

    Open claude.ai, pick Sonnet 5.5 and try it on the job you have half-finished. If the result is not good enough, move up to Opus. If the job repeats thousands of times, drop down to Haiku. Save Fable for when the others have given up.

    Sources

    Verified 17 November 2026.

    Gnurt

    Full-stack developer

    Discuss your project

    Comments

    No comments yet. Be the first!