OverpayingForAIPricing desk

Lesson 2 of 8 · 10 min read · Beginner

Fable, Opus, Sonnet and Haiku explained

Claude ships as a ladder of models at very different prices. Learn what each tier is for, how context windows work, and a one-line rule for picking the right one.

In this lesson you will

  • Place Fable, Opus, Sonnet and Haiku on a cost-versus-capability ladder
  • Pick a model for a task in under ten seconds
  • Understand context windows and why '1M tokens' is a ceiling, not a target

Anthropic names its models by tier, not by number. Each tier is a different trade-off between capability and price, and each generation (4.5, 4.6, 5, 5.1…) refreshes the ladder. In the app you pick from a model menu; in the API you pass a model id string. Either way, the choice is where most of your money goes.

Current Claude models — $ per 1M tokensLive from the OverpayingForAI catalogue · last verified 2026-09-04 · sorted by output priceInputOutputClaude Fable 5.11M context$10$50Claude Opus 4.81M context$5$25Claude Opus 51M context$5$25Claude Sonnet 4.61M context$3$15Claude Sonnet 51M context$2$10Claude Haiku 4.5200K context$1$5Batch, fast-mode and free variants are excluded. Prices change; the figure re-draws from the catalogue on every build.
Figure 1.The Claude ladder from the live catalogue. Output tokens (dark bars) cost five times input tokens on every tier, and each step up roughly doubles or more.

The four tiers

Claude model tiers at the time of writing
TierCurrent modelBest forContextRule of thumb
HaikuClaude Haiku 4.5High-volume, latency-sensitive work: classification, extraction, chat at scale200KCheapest — try it first for anything repetitive
SonnetClaude Sonnet 5The everyday workhorse: writing, coding, analysis, most agent steps1MDefault for 80% of tasks
OpusClaude Opus 5Hard reasoning, long autonomous coding sessions, research synthesis1MWhen Sonnet's answer is not good enough
FableClaude Fable 5.1Anthropic's most capable widely released model; the hardest problems where correctness beats cost1MTwice the Opus rate — earn it on a benchmark first

Older generations — Sonnet 4.6, Opus 4.8, Opus 4.7 — remain available on the API at their own rates and are sometimes cheaper. The catalogue lists them all on the Models & Providers page. For new work, start on the current generation: the newer models at lower effort settings usually beat the older ones at full effort, and you keep one price list to manage.

Context windows

Tokens and the context windowA token is roughly ¾ of an English word. The context window is the maximum tokens the model can hold at once — everything you send plus everything it writes.System prompt + instructionsConversation history + documentsNew answerfreeWhy it costs moneyEvery turn re-sends the whole history. A 40-message chat with a pasted PDF bills that PDF 40 times unless the provider caches it.“1M context” is a ceiling, not a target. Filling it on every request is the fastest way to overpay.rule of thumb: 1,000 tokens ≈ 750 words ≈ 1.5 pages
Figure 2.The context window holds your instructions, the whole conversation so far, any files, and the new answer. Every turn re-sends all of it.

Current Sonnet, Opus and Fable models accept up to a million tokens — roughly 750,000 words — in one request. Haiku 4.5 accepts 200,000. That is enough for entire codebases or hundreds of pages, and it is the feature that makes Claude popular for document-heavy work.

It is also the feature that quietly drains API budgets. A long conversation re-sends its whole history on every turn, so the fortieth message in a chat with a pasted 100-page PDF bills that PDF for the fortieth time. In the app this shows up as hitting your usage limit faster; on the API it shows up on the invoice. Lesson 8 covers prompt caching, which is the fix.

Thinking and effort

Current Claude models can *think* before answering — reason through a problem in private before writing the response. In the app this is the extended thinking toggle. On the API it is adaptive by default on the newest models and you steer it with an effort setting from low to max. Higher effort means better answers on hard problems and more output tokens billed. For routine tasks, lower effort on a current model is both cheaper and often better than an older model at full effort.

Knowledge check

You need to tag 50,000 support tickets by topic. Which model should you try first?

Lesson FAQ

Is Fable available to everyone?

Claude Fable 5.1 is Anthropic's most capable widely released model and is available on the API and in the app. A separate variant, Claude Mythos, is limited to approved organisations. For most buyers the choice is between Haiku, Sonnet and Opus.

Why do output tokens cost more than input tokens?

Generating text is more computationally expensive than reading it, and every provider prices it that way. On Claude the ratio is five to one, which is why asking for shorter answers is the cheapest optimisation there is.

Do older Claude models get cheaper over time?

Sometimes. Anthropic has cut prices on new generations rather than discounting old ones, so the current Sonnet is often cheaper than the previous Opus. Check the pricing tracker before assuming the old model is the bargain.

Finished reading?

Mark it done to track your progress through the course.

Compare, calculate, decide — for Claude

If our calculators helped you cut down on hidden AI wallet leaks, consider buying us a coffee. A tiny fraction of your savings keeps our pricing indexes updated daily.

Not sure which AI is cheapest for your use case? Find out in 30 seconds — no signup required.

AI cost intelligence

Stop overpaying for AI tools

Join the OverpayingForAI list for pricing updates, cheaper alternatives, and practical buying guidance.

Now tracking 50+ AI tools, models, platforms, subscriptions, coding tools, and automation products.

We use your email only for OverpayingForAI updates. Unsubscribe anytime.