Claude Haiku 4.5
What will your workload cost?
API charges depend on the amount you send, the answer you receive and how often you run the task. A chat subscription is a separate purchase.
Published catalog rates
$1 / million
Input text tokens
$5 / million
Output text tokens
Endpoint: anthropic. Host: Anthropic. No additional long-context price tier is listed in this catalog entry.
Check pricing source ↗Context capacity: 200,000 tokens. Maximum output: 64,000 tokens. Capacity is a technical limit, not a quality guarantee.
Estimate a real workload →Use a representative document
Upload or paste a typical document. Choose summary, extraction or translation; TokenMunch estimates the answer length automatically. For monthly budgets, include repeat calls and retries. Review the planning range before quoting a client.
Model prices alone cannot tell you whether extraction is accurate or a translation is acceptable. Test examples from your own workflow. Caching, batch processing, reasoning and external tools may change the final invoice.
Read all assumptions and exclusions →