Claude Sonnet 5 vs Claude Opus 5
Start with Sonnet 5. Upgrade to Opus 5 only when eval data proves you need it.
Claude Opus 5 is 5× more expensive than Sonnet 5 ($15 vs $3/M input). Both have 200K context. Opus 5 scores higher on complex reasoning tasks, but Sonnet 5 handles 85%+ of professional tasks at comparable quality. The right question is not "which is better?" but "what's the minimal quality level my task requires?"
Side-by-side pricing
| Claude Sonnet 5 | Claude Opus 5 | |
|---|---|---|
| Provider | Anthropic | Anthropic |
| Input price/1M | $3.00 | $5.00 |
| Output price/1M | $15.00 | $25.00 |
| Cache read/1M | $0.30 | $0.50 |
| Context window | 200K | 200K |
| Vision | ✓ | ✓ |
| Tool use | ✓ | ✓ |
| Prompt caching | ✓ | ✓ |
| Speed tier | Standard | Slow |
Prices verified 2026-08-20. See changelog for history.
Winner by task
Calculate costs for each model
Frequently Asked Questions
When is Claude Opus 5 worth the premium?
When your task has measurable quality signals (evals, user ratings) that show Sonnet 5 failing at a rate that matters to your business. Examples: legal contract review where errors have material cost, complex code review where Sonnet misses critical bugs. Not worth it for: customer support, summarization, classification.
Prices verified 2026-08-20. See methodology for calculation details.