{{ p.category }}

{{ t }}

{{ p.title }}

{{ p.excerpt }}

U{{ author.name }} {{ p.readMins }} min read

{{ p.category }}

{{ p.title }}

{{ mo.sub }}

{{ m.tag }}

{{ m.t }}

{{ m.d }}

{{ m.left.t }}

✕{{ x.t }}

{{ m.right.t }}

✓{{ x.t }}

{{ m.big }}{{ m.label }}

{{ x.t }}

{{ m.label }}

✓{{ x.t }}

{{ x.t }}

{{ m.label }}
{{ x.t }}
{{ m.status }}

{{ m.label }}

{{ x.t }}{{ x.n }}

{{ x.t }}

{{ m.label }}{{ m.stepTxt }}

{{ x.t }}

{{ m.btn }}

{{ x.t }}

{{ m.q }}

{{ m.a }}

{{ x }}

{{ k.n }}

{{ k.v }}

{{ m.label }}

{{ m.status }}

{{ m.label }}

{{ st.n }}{{ st.t }}

{{ m.hub }}

{{ nd.t }}

AaAaAaAa

ButtonInput fieldTag

THE SHORT ANSWER

{{ p.answer }}

KEY TAKEAWAYS

✓{{ k }}

RELATED SERVICES

{{ b.t }}

{{ g.t }}{{ g.t }}

{{ g.t }}

{{ b.label }}:{{ b.t }}

{{ c.ini }}

{{ c.t }}

{{ c.d }}

{{ s.n }}

{{ s.t }}

{{ s.d }}

{{ h }}
{{ c.t }}

{{ b.alt }}

{{ g.alt }}

{{ b.title }}

{{ c.t }}

{{ c.s }}

{{ i }}

{{ c.t }}{{ c.s }}

{{ b.caption }}

{{ s.v }}

{{ s.l }}

{{ s.s }}

{{ b.caption }}

{{ b.title }}

{{ b.unit }}

{{ r.l }}{{ r.d }}

{{ b.caption }}

{{ b.title }}

A{{ b.a.t }}

{{ i }}

B{{ b.b.t }}

✓{{ i }}

{{ b.caption }}

{{ b.title }}

{{ t.d }}

{{ t.t }}

{{ b.caption }}

“

{{ b.t }}

{{ b.by }}

{{ b.title }}

{{ r.d }}

{{ r.l }}

{{ b.caption }}

{{ b.title }}

{{ m.n }}

{{ b.lo }}{{ b.hi }}

{{ m.n }}

{{ m.l }}

{{ m.d }}

{{ b.caption }}

{{ b.cs.client }} case study

Case study · {{ b.cs.client }} {{ b.cs.title }} {{ b.cs.text }}

{{ f.v }}{{ f.l }}

View case study →

SOURCES AND FURTHER READING

Frequently asked questions

{{ f.a }}

U

Written by {{ author.name }}

{{ author.role }}

Keep reading

More articles from across design, AI and marketing.

All articles

{{ r.category }} · {{ r.readMins }} min{{ r.title }}

Special Offer

25% Off Your Startup’s First Project

No Obligation
Quick Response
Exclusive for Startups

Claude Opus vs Sonnet vs Haiku: which Claude model should you use?

Use Claude Opus 5.5 for the hardest work, such as long agentic coding and complex analysis; Claude Sonnet 5.5 for most everyday writing, coding and analysis at half Opus’s price; and Claude Haiku 5.5 for high-volume, fast tasks at a fraction of the cost. On the API, Opus costs $4/$20, Sonnet $2/0 and Haiku from $0.10/$0.50 per million input/output tokens.

How do Opus, Sonnet and Haiku compare?

	Claude Opus 5.5	Claude Sonnet 5.5	Claude Haiku 5.5
Released	22 September 2026	28 September 2026	7 October 2026
Positioning	Long-running agentic coding and knowledge work	Fast, capable everyday model	Fastest, for high-volume tasks
API price (input / output per million tokens)	$4 / $20	$2 / 0	From $0.10 / $0.50
Context window	1M tokens	1M tokens	1M tokens
Speed	Moderate	Fast	Fastest

Specifications come from Anthropic’s Claude Opus 5.5 model overview. Haiku’s starting price applies to prompts up to 100,000 tokens, with higher rates above that, as noted in this pricing breakdown.

When should you use Claude Opus?

Opus is Anthropic’s most capable mainstream model and the one its documentation suggests starting with for demanding workloads. It is best for multi-step agentic tasks, large codebase changes, complex research and analysis, and work where a wrong answer is expensive. It is slower and costs twice as much per token as Sonnet, so it pays off where quality differences are large.

When should you use Claude Sonnet?

Sonnet is the practical default for most teams. It handles writing, editing, coding, data analysis and document review well, responds faster than Opus and costs half as much per token. For many everyday tasks the difference in output quality between Sonnet and Opus is small. Start with Sonnet and move a task up to Opus only when results fall short.

When should you use Claude Haiku?

Haiku is built for speed and volume: classifying support tickets, extracting fields from documents, moderating content, routing requests and powering responsive chat features. At a starting price of $0.10 per million input tokens it is about a fortieth of Opus’s input price, which makes it the natural choice for tasks that run thousands or millions of times.

What about Fable and Mythos?

Anthropic also lists Claude Fable 5.1 as a top-tier model for cases where capability matters more than cost, priced well above Opus. Claude Mythos is a separate series that Anthropic has offered only to selected organisations for security work, under a programme called Project Glasswing, rather than as a general product. For most business use, the choice is between Opus, Sonnet and Haiku.

Which models are available in the Claude apps?

In the Claude apps, paid plans such as Pro, Max, Team and Enterprise include Opus, while the free plan offers Sonnet and Haiku. API customers can call any model directly and pay per token. If you are comparing Claude with other assistants, our guide to ChatGPT vs Claude vs Gemini covers the wider choice.

For an overview of everything Claude offers, see what is Claude.

Which Claude model fits each task?

Task	Suggested model
Large refactors and long agentic coding sessions	Opus
Complex research reports and strategy analysis	Opus
Everyday writing, editing and summaries	Sonnet
Feature-level coding and code review	Sonnet
Customer-facing chat that must respond quickly	Haiku or Sonnet
Ticket triage, tagging and data extraction at scale	Haiku
Sub-agents doing simple steps inside a larger workflow	Haiku

For coding, our comparison of Claude Code, Cursor and Codex explains how the models are used inside coding tools.

How can you control Claude costs?

Route by difficulty: Send each task to the smallest model that does it well.
Use prompt caching: Reuse long instructions or documents across requests at a lower cached rate.
Batch non-urgent work: The Batch API halves prices for jobs that can wait.
Keep prompts lean: Long prompts cost more on every call, and Haiku charges more above 100,000 tokens.
Measure quality and spend: Track results and cost per task, and revisit routing when new models launch.

How should you test which model is right?

Take 20 to 50 real examples of a task, run them through Sonnet and one neighbouring tier, and score the outputs blind. If Haiku scores as well as Sonnet, switch and save. If Sonnet misses important details that Opus catches, escalate only that task. Repeat the test when Anthropic releases new versions, as the gaps between tiers change.

What mistakes do teams make when choosing a Claude model?

Using Opus for everything and paying for capability simple tasks do not need
Using Haiku for complex reasoning and blaming the model for weak answers
Comparing models on one or two examples instead of a representative set
Ignoring speed: a slower, smarter model can make a chat feature feel broken
Forgetting to re-test after new releases, when a cheaper tier may now be good enough

A simple routing rule, such as Haiku for classification, Sonnet for drafting and Opus for multi-step agent work, usually captures most of the savings without complex infrastructure.

Above Opus sits the Mythos-class tier; our guide to Claude Fable and Mythos explains it, and the Claude API guide covers pricing and a first request.

If you need to run models on your own infrastructure, our guide to open-weight AI models compares DeepSeek, Qwen and Llama.

How UIUX Studio helps

We design and build Claude-powered features and agents, including model routing, caching and evaluation so costs stay predictable. Talk to our Agentic AI & Automation team, or read how we approach AI in SaaS products.

Frequently asked questions

Which Claude model is the best?

Opus 5.5 is the most capable mainstream model, but Sonnet 5.5 is the better default for most everyday work.

Is Haiku good enough for customer chat?

Often, yes, for straightforward questions. Escalate complex conversations to Sonnet.

Do all three have the same context window?

Yes. Opus 5.5, Sonnet 5.5 and Haiku 5.5 each support up to 1 million tokens.

Can I use Opus on the free plan?

No. Opus is included in paid plans; the free plan offers Sonnet and Haiku.

How much cheaper is Haiku than Opus?

On the API, Haiku’s starting input price is about a fortieth of Opus’s.

Related articles