All 4 Anthropic models we track, priced per 1 million tokens in US dollars, taken from Anthropic’s own pricing page and last checked 2026-10-25. Click a date to open the source.
| Model | Input | Cached input | Output | Checked |
|---|---|---|---|---|
| Haiku 5.5 Rate for prompts up to 100K tokens; longer prompts cost $0.50 in / $2.50 out. | $0.1 | $0.01 | $0.5 | 2026-10-25 |
| Sonnet 5.5 | $2 | $0.1 | $10 | 2026-10-25 |
| Opus 5.5 | $4 | $0.2 | $20 | 2026-10-25 |
| Fable 5.1 Flagship reasoning model. | $10 | $0.25 | $50 | 2026-10-25 |
US dollars per 1,000,000 tokens. Notes under a model name flag long-context surcharges, off-peak rates and announced price changes.
Per-token prices are hard to picture, so here are two common workloads priced for a 30-day month. The cached column assumes 70% of each chat request’s input is a repeated prefix (system prompt and earlier turns) billed at the cached rate.
| Model | Chatbot / month | Chatbot, cached / month | Documents / month |
|---|---|---|---|
| Haiku 5.5 | $12.00 | $8.22 | $18.75 |
| Sonnet 5.5 | $240 | $160 | $375 |
| Opus 5.5 | $480 | $320 | $750 |
| Fable 5.1 | $1,200 | $791 | $1,875 |
Plug in your own traffic in the cost calculator, or line these models up against every other provider in the comparison table.
Minimum prompt lengths, cache lifetimes and worked examples for every provider are in our guide to prompt caching and batch discounts.
Haiku 5.5, at $0.1 per million input tokens and $0.5 per million output tokens.
Fable 5.1, at $10 per million input tokens and $50 per million output tokens.
Yes. 4 of the 4 models listed have a cached-input rate, 90–98% below their normal input price.
Roughly 750,000 English words: a token is about four characters, or three quarters of a word. Code and most other languages use more tokens per word.
OpenAI API pricingGemini API pricingGrok API pricingDeepSeek API pricing
Every price comes from the provider’s own pricing page: claude.com/pricing