Opus vs Sonnet vs Haiku:
which Claude model?
How to choose between Claude Opus, Sonnet and Haiku by task and budget, with live prices side by side and a quick test to settle it on your own data.
Claude comes in three sizes. Haiku is the small, fast, cheap one; Sonnet is the balanced middle; Opus is the largest and most capable. Newer generations of each keep arriving, so the names are more stable than the numbers, and the useful question is not "which is best" but "which is the cheapest that still gets my task right." This guide gives a default for common jobs and a ten-minute test that replaces opinion with your own data.
The three sizes at a glance
| Haiku | Sonnet | Opus | |
|---|---|---|---|
| Speed | Fastest | Fast | Slowest |
| Price | Lowest | Middle | Highest |
| Strength | High-volume, simple and structured tasks | Most everyday work, coding, agents | Hardest reasoning, long multi-step work |
| Typical use | Classification, extraction, routing, short replies | Chatbots, coding assistants, summaries, tool use | Complex analysis, large refactors, planning |
Live prices
| Model | Context | Input | Output | Cache read |
|---|---|---|---|---|
| Claude Fable 5.1 claude-fable-5.1 | 1M | $5.00 | $25.00 | $0.13 |
| Claude Opus 5.5 claude-opus-5.5 | 1M | $2.00 | $10.00 | $0.10 |
| Claude Sonnet 5.5 claude-sonnet-5.5 | 1M | $1.00 | $5.00 | $0.05 |
| Claude Haiku 5.5 claude-haiku-5.5 | 1M | $0.05 | $0.25 | $0.005 |
| Claude Fable 5 claude-fable-5 | 1M | $5.00 | $25.00 | $0.50 |
| Claude Opus 5 claude-opus-5 | 1M | $2.50 | $12.50 | $0.25 |
| Claude Sonnet 5 claude-sonnet-5 | 1M | $1.00 | $5.00 | $0.10 |
| Claude Opus 4.8 claude-opus-4.8 | 1M | $2.50 | $12.50 | $0.25 |
| Claude Opus 4.7 claude-opus-4.7 | 1M | $2.50 | $12.50 | $0.25 |
| Claude Opus 4.6 claude-opus-4.6 | 1M | $2.50 | $12.50 | $0.25 |
| Claude Sonnet 4.6 claude-sonnet-4.6 | 1M | $1.50 | $7.50 | $0.15 |
| Claude Opus 4.5 claude-opus-4.5 | 200K | $2.50 | $12.50 | $0.25 |
| Claude Sonnet 4.5 claude-sonnet-4.5 | 1M | $1.50 | $7.50 | $0.15 |
| Claude Haiku 4.5 claude-haiku-4.5 | 200K | $0.50 | $2.50 | $0.05 |
USD per 1M tokens, live from the catalog.
Read the input and output columns, not just one: a model with cheap input and expensive output is a poor fit for a task that writes a lot.
A default for common jobs
Haiku. These tasks are short and structured, and volume is high, so price per call dominates.
Sonnet, or Haiku if the answers are short and grounded in retrieved text.
Sonnet as the everyday model, Opus when a task keeps failing or needs careful planning. In Claude Code you can set both: the main model and a small model for background steps.
Opus.
Settle it with a ten-minute test
- Collect 20 real inputs and write down what a good answer looks like.
- Run all 20 on the smallest model. Note which fail.
- Run the failures on the next size up. Repeat.
- Multiply the winning model's price by your monthly volume in the calculator.
The result is usually a split: the small model handles most traffic, and the bigger one is called only for the hard cases. Routing by a cheap first pass, or by task type, often cuts the bill by more than any prompt trick.
What changes the cost besides the model
Output tokens cost several times input tokens, extended thinking is billed as output, and cached input is a small fraction of the normal price. A smaller model with long thinking can cost more than a larger one answering directly, so compare total tokens per task, not only the price per million. Claude API pricing explains each part, and the cheapest way to use the Claude API lists the other levers.
Frequently asked questions
Is Opus worth the extra cost?+
Only for work that the smaller models get wrong. If Sonnet passes your test cases, Opus is paying for capability you don't use. If it fails them, the larger model is cheaper than the retries.
Which Claude model is best for coding?+
Sonnet is the usual default for coding assistants and agents; step up to Opus for large, ambiguous or failing tasks. Test both on a real task from your repository.
Which Claude model is the cheapest?+
Haiku, the smallest size, has the lowest prices in the table above.
Do all Claude models support the same features?+
Prompt caching, tool use, vision and extended thinking are broadly available, but check each model's page for its context window and supported options before relying on one.
How do I switch models?+
Change the model string in your request; nothing else changes. The ids are on the model list, and the quickstart shows a first request. Create an account to try them on a prepaid balance at half the list price.