Per-million-token input and output prices for the three Anthropic API tiers.
Apply prompt caching to drop input cost to 10% on cache hits, and use the Batch API for a further 50% discount when latency is flexible.
CLAUDE.md