Live Claude prices
Read from the live catalog when this page opens.
Three ways to pay less for Claude
- Use the smallest model that works: Haiku-class models cost a fraction of Opus-class models and handle most routine work.
- Use prompt caching: chat apps resend the conversation every turn, and cached input is billed at the lower cached rate on supported models.
- Compare a plan with pay-as-you-go: heavy users can compare a Plus subscription with their monthly usage using the price calculator.
Anthropic or OpenAI format
Claude models that list the Anthropic format work with tools such as Claude Code. Some list only the OpenAI-compatible format, so check each model page first. Both formats use the same account and key; see compatibility.
Troubleshooting
| Symptom | What it usually means | What to do |
|---|---|---|
| Cached input shows a dash | The catalog does not list a cached input price for that model. | Budget at the normal input price. |
| Actual spend differs from an estimate | Output length, cache hits and request mix differ from the estimate. | Estimate with a representative request, then check usage in the console. |
Frequently asked questions
Are these prices hard-coded?
No. The table is loaded from the live catalog every time the page opens.
When does the cached input price apply?
Only when the model supports caching and a request actually reuses cached context.