Hong Life CalculatorsHow it works
Cost per request = (uncached input tokens × input price + cached input tokens × cached-input price + output tokens × output price) ÷ 1,000,000, with prices in US$ per 1 million tokens. The cost of requests using Message Batches API is multiplied by the provider’s discount factor (0.5). Day = request cost × requests per day; month = day × days per month; year = month × 12. “Saved” compares with the same traffic at the plain input and output price. Cached input = cache hit price. Writing a prompt to the 5-minute cache costs 1.25x the input price (1-hour cache: 2x). Claude 4.6 and later (except Haiku 5.5) bill the whole 1M context at standard rates; Haiku 5.5 prompts over 100,000 tokens cost more. Taxes, tools (web search, code execution), image or audio tokens, long-context surcharges and cache-write fees are not included.
Anthropic API prices (US$ per 1M tokens, checked 10 Nov 2026)
| AI model | Input price | Cached-input price | Output price | Message Batches API input / output |
|---|---|---|---|---|
| Claude Fable 5.2 | 8 | 0.2 | 40 | 4 / 20 |
| Claude Mythos 5.2 | 8 | 0.2 | 40 | 4 / 20 |
| Claude Opus 5.5 | 4 | 0.2 | 20 | 2 / 10 |
| Claude Sonnet 5.5 | 2 | 0.1 | 10 | 1 / 5 |
| Claude Haiku 5.5 (≤100K prompt) | 0.1 | 0.01 | 0.5 | 0.05 / 0.25 |
| Claude Opus 5 | 5 | 0.5 | 25 | 2.5 / 12.5 |
| Claude Sonnet 5 | 2 | 0.2 | 10 | 1 / 5 |
| Claude Opus 4.8 | 5 | 0.5 | 25 | 2.5 / 12.5 |
| Claude Sonnet 4.6 | 3 | 0.3 | 15 | 1.5 / 7.5 |
| Claude Haiku 4.5 | 1 | 0.1 | 5 | 0.5 / 2.5 |
⚠ Estimates for information only — not financial advice.
FAQ
How much does prompt caching save on Claude?
A cache hit is billed at 5% to 10% of the normal input price, depending on the model (see the table). Writing the prompt into the cache costs 1.25× the input price once per 5 minutes, so caching pays off when the same long prefix is reused several times.
Does Claude use more tokens for the same text?
Anthropic states that Claude 4.7 and later models use a tokenizer that produces about 30% more tokens for the same text. If you measured token counts with an older model, add about 30% to the input and output tokens.
Checked 10 Nov 2026. Anthropic official API pricing page (https://platform.claude.com/docs/en/about-claude/pricing), read on 10 Nov 2026. Batch API: 50% off input and output tokens.