What you will be able to do
- Explain the single trade-off that separates the Haiku, Sonnet and Opus tiers
- Describe what each tier is like to use: reasoning depth, speed and how much quota it uses
- Read a tier comparison and identify which tier is fastest, cheapest and deepest-reasoning
Key concept
Tier as a trade-off — Haiku, Sonnet and Opus belong to one model family. Each one balances reasoning depth against speed and cost differently. Moving up a tier gets you more reasoning, but responses are slower and each piece of work costs more.
1.One family, one trade-off
Many people new to Claude assume the tier names work like product editions: a basic one, a standard one and a premium one, each with a different feature list. That picture is misleading. Anthropic's lineup is one family of models, each balancing intelligence, speed and cost differently. Choosing well means matching the model to the job.
The key word is *balancing*. You do not get every benefit from one tier. A tier that reasons more deeply takes longer to respond and costs more per piece of work. A tier that answers almost instantly and costs little gives up some reasoning depth. None of the three is simply "best". Each sits at a different point on the same curve.
The basic capabilities are also shared. Anthropic's models overview says that all current models accept text and images, produce text, work across languages, and can use tools. Moving from Haiku to Opus does not unlock the ability to read a screenshot or call a tool. It changes how much reasoning the model brings to the task, how quickly it answers, and how much of your budget each answer uses.
Checkpoint 1 of 4· Check yourself
A colleague says they need Opus because their task involves reading an image and the cheaper tiers cannot do that. What is the accurate response?
Image input, vision, multilingual capability and tool use are common to all current models. The tiers differ in reasoning depth, speed and cost, not in these basic capabilities.
“All current models support text and image input, text output, multilingual capabilities, vision, and tool use.”Source: platform.claude.com
2.What each tier is like to use
Seen from the user's side, each tier has a recognisable character. Anthropic's help centre describes them plainly.
Opus is the deep-reasoning tier. It offers deeper reasoning for harder problems. The help centre's examples are large cross-cutting refactors, difficult debugging and architectural decisions. That depth has a price: Opus uses meaningfully more of your quota, and the same guidance suggests switching to Sonnet for routine work. Anthropic's model guide calls Opus its powerful model class for reasoning-intensive enterprise tasks.
Sonnet is the everyday workhorse. The help centre sums it up in one line: fast, capable and cost-efficient. Anthropic calls Sonnet its versatile model class for everyday tasks, with a balance of performance, cost and speed for the widest range of general-purpose work. In practice, most work belongs on Sonnet by default.
Haiku is the speed-and-economy tier. It is the fastest and cheapest option, well suited to quick lookups, simple edits or high-volume scripted runs. Anthropic also describes Haiku models as designed for high-frequency workloads where latency and cost matter.
Checkpoint 2 of 4· Match them up
Match each tier to how it behaves in everyday use
Tap a term, then the definition that fits it.
The help centre gives each tier a distinct character. Opus trades quota for reasoning depth, Sonnet is the efficient general choice, and Haiku gives up some depth for speed and low cost.
“Haiku is the fastest and cheapest option, well suited to quick lookups, simple edits, or high-volume scripted runs.”Source: support.claude.com
Checkpoint 3 of 4· Exam question
A voice assistant application requires the lowest possible response latency among current Claude models so that spoken exchanges feel natural and immediate. Which model tier has the fastest comparative latency?
Correct answer: A — Claude Haiku, since it is described as having the fastest comparative latency of the current model lineup
- A. Correct. In Anthropic's model comparison, Haiku is listed with the fastest comparative latency among the current-generation models.
- B. Sonnet is described as fast, but the comparison table ranks Haiku's comparative latency ahead of Sonnet's.
- C. Greater intelligence does not translate into lower latency; Opus is listed with moderate comparative latency, slower than both Sonnet and Haiku.
- D. The models differ noticeably in comparative latency, not just in output quality, which is precisely why latency is one of the criteria for choosing among them.
3.The tiers side by side
Anthropic's models overview turns these descriptions into figures. The table below takes the current model in each tier and shows how the overview positions it, its comparative latency and its published price per million tokens. The exact numbers matter less than the pattern: each step up the ladder makes responses slower and roughly doubles the price.
| Tier (current model) | How Anthropic positions it | Comparative latency | Price per million tokens (input / output) |
|---|---|---|---|
| Claude Haiku 4.5 | The fastest model with near-frontier intelligence | Fastest | $1 / $5 |
| Claude Sonnet 5.5 | The best combination of speed and intelligence | Fast | $2 / $10 |
| Claude Opus 5.5 | For long-running agentic coding and knowledge work | Moderate | $4 / $20 |
Two details in the table are easy to miss. First, Haiku is described as having *near-frontier intelligence*, not as a toy model. Its edge is speed and price, and it is a capable model in its own right. Anthropic's model guide recommends it where the lowest latency and price matter, and notes that it supports extended thinking. Second, Opus is not the fastest at anything. Its latency is listed as moderate because the extra reasoning takes time.
The lineup also includes a class above Opus, called Fable, for the most demanding long-running work. On the Claude apps, though, the choice you will usually face is between the three tiers this lesson covers. Version numbers will move on, because exact model names, versions and availability change over time. The ordering of trade-offs is the part that stays stable.
Checkpoint 4 of 4· Check yourself
Based on the models overview, which statement about the three tiers is accurate?
The overview lists Haiku as the fastest tier and describes it as having near-frontier intelligence. Opus has moderate latency, and each tier up costs more per token.
“The fastest model with near-frontier intelligence”Source: platform.claude.com
Exam traps
Each one states something that sounds right. Open it to see what is actually true.
1.Haiku is a stripped-down model that can only do trivial work, so it is never a serious option.Why is that wrong?
Haiku is the fastest and cheapest tier, and Anthropic still describes it as having near-frontier intelligence. It is chosen for speed and economy, not because nothing else is possible.
Covered in The tiers side by side
2.Sonnet is a compromise you settle for when Opus is unavailable.Why is that wrong?
Sonnet is the efficient default for routine work. The help centre itself suggests switching from Opus to Sonnet for routine tasks.
Covered in What each tier is like to use
Sources
Every claim above is drawn from one of these pages, quoted as it was written on the date shown.
- 1.https://platform.claude.com/docs/en/models/overviewOfficial docs
“All current models support text and image input, text output, multilingual capabilities, vision, and tool use.”
↩︎ One family, one trade-off“The fastest model with near-frontier intelligence”
↩︎ Exam trap 1 - 2.https://claude.com/blog/claude-models-explained-choosing-the-best-model-for-your-use-caseSecondary source
“each balancing intelligence, speed, and cost differently. Choosing well means matching the model to the job.”
↩︎ One family, one trade-off“Opus is our powerful model class for reasoning-intensive enterprise tasks.”
↩︎ What each tier is like to use“Sonnet is our versatile model class for everyday tasks.”
↩︎ What each tier is like to use“Haiku models are designed for high-frequency workloads where latency and cost matter.”
↩︎ What each tier is like to use“how hard a problem they can reliably carry, and what that capability costs in price and speed”
↩︎ Prediction - 3.
“Opus offers deeper reasoning for harder problems”
↩︎ What each tier is like to use“It is fast, capable, and cost-efficient.”
↩︎ What each tier is like to use“Exact model names, versions, and availability change over time.”
↩︎ The tiers side by side“It uses meaningfully more of your quota, so consider switching to Sonnet for routine work.”
↩︎ Exam trap 2“Haiku is the fastest and cheapest option, well suited to quick lookups, simple edits, or high-volume scripted runs.”
↩︎ Checkpoint - 4.
“The lowest latency and price, with extended thinking”
↩︎ The tiers side by side“Choosing a Claude model means balancing capabilities, speed, and cost.”
↩︎ Key concept
Also cited
“The fastest model with near-frontier intelligence”
↩︎ Checkpoint