One of the first questions when building with Claude is which model to use. Anthropic offers three tiers — Haiku, Sonnet, and Opus — each with a different balance of speed, capability, and cost. Here's how to think about the choice.
The three tiers
Haiku is the fastest and cheapest. It's designed for high-volume tasks where speed matters and the job is relatively simple — classifying text, extracting structured data from a template, answering short factual questions. If you're processing thousands of documents or need sub-second responses, start here.
Sonnet is the sweet spot for most applications. It's significantly more capable than Haiku but still fast enough for interactive use. This is what I reach for by default — it handles complex reasoning, code generation, and nuanced writing well, without the cost of Opus.
Opus is the most capable model in the family. Use it when the task genuinely requires it — complex multi-step reasoning, difficult coding problems, tasks where quality matters more than speed or cost. It's slower and more expensive, so it's worth being deliberate about when you need it.
How to choose
Default to Sonnet. It handles the vast majority of real-world tasks well. Only move up or down once you have a reason to.
Move to Haiku when:
- You're running high volumes (many calls per minute)
- Latency is critical
- The task is simple and well-defined (classification, extraction, short Q&A)
- Cost is a constraint
Move to Opus when:
- Sonnet's output quality isn't good enough
- The task involves complex reasoning or judgement
- You're doing a low-volume, high-stakes task (generating a report, writing production code)
A practical pattern
A common approach in production systems is to use different models for different steps in the same pipeline. A cheap Haiku call might handle initial triage or routing, Sonnet handles the main task, and Opus is only called for edge cases that require escalation. You get most of the capability at a fraction of the cost.
Current models
At the time of writing, the current Claude 4 lineup is:
| Model | ID | Best for |
|---|---|---|
| Haiku 4.5 | claude-haiku-4-5-20251001 | Speed, volume, simple tasks |
| Sonnet 4.6 | claude-sonnet-4-6 | General use, most applications |
| Opus 4.7 | claude-opus-4-7 | Complex reasoning, highest quality |
Model versions change — check the Anthropic docs for the latest IDs before hardcoding them.