Claude Opus vs Sonnet vs Haiku: Which Model Should You Use?
Anthropic's Claude models come in tiers — Opus, Sonnet and Haiku, plus Fable at the top. What each is good at, what they cost on the API, and a simple rule for choosing for coding, chat and features inside your own app.
Anthropic's Claude models come in families named after literary forms. The names describe size and price, not age: a new version of each arrives every few months. Choosing the right one is the biggest single lever on cost and speed, whether you're coding with Claude Code or calling the API from your own app.
Model names and prices below are from Anthropic's model documentation in October 2026. New versions arrive often; the tier logic stays the same.
The tiers in one sentence each
- Haiku — small, fast and cheap. Great for simple, high-volume jobs.
- Sonnet — the everyday workhorse. Fast and capable enough for most coding and most app features.
- Opus — the strongest of the three for hard reasoning, long agentic tasks and tricky code.
- Fable — Anthropic's most capable widely released model, priced above Opus, for the most demanding long-horizon work.
Current versions and API prices
Prices are per million tokens (what's a token?):
| Model | Model ID | Input | Output | Context window |
|---|---|---|---|---|
| Claude Fable 5.1 | claude-fable-5-1 |
$10 | $50 | 1M tokens |
| Claude Opus 5.5 | claude-opus-5-5 |
$4 | $20 | 1M tokens |
| Claude Sonnet 5.5 | claude-sonnet-5-5 |
$2 | $10 | 1M tokens |
| Claude Haiku 4.5 | claude-haiku-4-5 |
$1 | $5 | 200K tokens |
Notice two things. Output tokens cost five times input tokens on every model, so long answers cost more than long questions. And the gap between tiers is roughly 2× at each step — Opus costs twice Sonnet, which costs twice Haiku.
For the full picture, including caching and batch discounts, see Claude API pricing explained.
What each tier is best at
Haiku
- Classifying, tagging and routing (is this support ticket about billing or a bug?).
- Extracting fields from text into JSON.
- Short summaries and rewrites.
- Sub-tasks inside a larger agent where speed matters more than depth.
Sonnet
- Most coding: features, bug fixes, tests, refactors with a clear goal.
- Chatbots and assistants in your app.
- Writing and editing.
- Agents with tools, where it's fast and reliable.
Opus
- Architecture and design decisions across a large codebase.
- Bugs where the cause isn't obvious and needs investigation.
- Long, multi-step agent runs where staying on track matters.
- Anything where a mistake is expensive.
Fable
- The hardest, longest-running tasks: multi-hour autonomous work, deep research, problems other models fail at.
- Worth its price when one correct answer replaces several failed attempts with a cheaper model.
A simple rule for choosing
- Start with Sonnet. It's the right answer far more often than not.
- Move down to Haiku for simple, repetitive, high-volume work where Sonnet is overkill.
- Move up to Opus when Sonnet gets it wrong, or the task is genuinely hard.
- Reach for Fable when Opus isn't enough and the outcome justifies the cost.
Judge cost per finished task, not per request. A cheap model that needs four attempts can cost more than a capable one that gets it right first time.
In Claude Code
Switch with /model opus, /model sonnet or /model haiku. The default on paid plans is currently Opus 5.5. Anthropic's own guidance: Sonnet handles most coding well and costs less; save Opus for complex reasoning. Using Sonnet for routine work also makes usage limits last longer. See how to change the model in Claude Code.
In your own app
Pick per feature, not per app. A support chatbot might use Sonnet, the ticket classifier behind it Haiku, and a once-a-day report generator Opus. The model is just a string in your API call, so changing it is easy — test with real examples before and after. See how to add an AI chatbot to your app.
There's also an effort setting on current models that controls how much the model thinks before answering. Lower effort on an easy task saves tokens without changing model.
The summary
- Haiku: fast and cheap. Sonnet: everyday default. Opus: hard problems. Fable: the hardest.
- Each tier costs roughly twice the one below it.
- Start with Sonnet, move down for volume, move up for difficulty.
- Measure cost per completed task, not per request.
EasySpawn runs Claude Code with your own Claude subscription or API key on a persistent server, so you choose the model per task and your app's AI features run on the same server with their keys kept server-side. See how it works or join the waitlist.
Related: What Is an LLM? · Claude Pro vs Max vs API Key · What Is a Context Window? · Prompt Caching Explained
Keep reading
OpenAI API vs Claude API: Which Should Your App Use?
Both APIs let your app send text (and images) to a model and get an answer back. How they differ in request shape, models, tool calling, structured output, caching and pricing — and why many apps keep the choice swappable instead of picking forever.
Claude Code vs Gemini CLI: Which Terminal Coding Agent Should You Use?
Claude Code and Google's Gemini CLI are both AI coding agents that live in your terminal. How they compare on models, cost, free tiers, extensibility, safety controls and day-to-day workflow — and how to pick.