🟣 Claude Token Calculator
What a prompt costs on Anthropic's Claude models. Claude's distinguishing feature is its cache pricing: a cache write costs 1.25x the input rate, but a cache read costs as little as 0.025x, which changes the economics of any repeated prompt.
Claude's cache economics, and the Haiku threshold that catches people out
Anthropic's pricing has two features that matter more than the headline rates. The first is an unusually deep cache discount, with reads costing as little as 2.5% of the input rate. The second is a prompt-length threshold on Haiku that multiplies the price fivefold once crossed, and it is easy to cross without noticing.
Current prices
| Model | Input | Output | Cache read |
|---|---|---|---|
| Claude Fable 5.1 | $10 | $50 | $0.25 |
| Claude Opus 5.5 | $4 | $20 | $0.2 |
| Claude Sonnet 5.5 | $2 | $10 | $0.1 |
| Claude Haiku 5.5 | $0.1 | $0.5 | $0.01 |
| Claude Haiku 5.5 (>100K) | $0.5 | $2.5 | $0.05 |
| Claude Sonnet 4.5 | $3 | $15 | $0.3 |
How the cache multipliers work
A cache write costs 1.25 times the input rate for a five-minute lifetime, or twice the rate for an hour. A cache read costs a fraction: 0.1x on most models, 0.05x on Sonnet 5.5, and 0.025x on Fable 5.1. Because the write premium is small and the read discount is enormous, caching pays for itself on the second or third use and is close to free after that.
The 100,000-token Haiku cliff
Haiku 5.5 costs $0.10 per million input tokens for prompts up to 100,000 tokens and $0.50 above it — a fivefold jump applied to the whole request, not just the excess. The threshold counts every input token including cache reads, so an application with a cached 90,000-token document can cross it by adding a long user message. If you are near the line, measure rather than assume.
Batch processing halves everything
The Batch API takes 50% off both input and output on every model, with results returned within hours instead of seconds. For evaluation runs, bulk classification, content generation reviewed later, or anything on a schedule, this is free money and a trivial change.
Choosing between the tiers
Fable sits at the top and costs a hundred times Haiku on input. Sonnet 5.5 at $2 and $10 is the usual production default, with Haiku handling classification and extraction beneath it. The previous-generation Sonnet 4.5 remains available at $3 and $15, which is 50% dearer than Sonnet 5.5 for an older model — worth checking if you pinned a version and never revisited it.
Prices come from Claude pricing and were checked on 2026-10-26. AI pricing changes frequently — confirm against the provider before committing to a budget.
Once you have this figure, the ai token calculator is where it usually goes next. the chatgpt token cost is the related one people reach for afterwards.
Frequently Asked Questions (FAQs)
How much does the Claude API cost?
On 2026-10-26 it ran from $0.10 per million input tokens on Haiku 5.5 to $10 on Fable 5.1, with output between $0.50 and $50. Sonnet 5.5 sits in the middle at $2 and $10.
How much does Claude's prompt caching save?
Cache reads cost 0.1x the input rate on most models, 0.05x on Sonnet 5.5 and 0.025x on Fable 5.1 — a saving of 90% to 97.5%. A cache write costs 1.25x for five minutes or 2x for an hour, so it pays back after two or three reuses.
What is the Haiku 100K threshold?
Prompts over 100,000 input tokens are billed at five times the standard Haiku rate, and the higher rate applies to the entire request. Cache reads count towards the threshold, so a cached long document plus a long message can cross it unexpectedly.
Is the Batch API worth using?
Yes, wherever latency does not matter. It is a flat 50% off both input and output, with results in hours rather than seconds, and it applies to every model.
Which Claude model should I use?
Sonnet 5.5 is the common production default at $2 and $10. Use Haiku for classification, extraction and routing, and reserve Opus or Fable for work where the quality difference is demonstrable and worth four to ten times the price.
Is Sonnet 4.5 cheaper than Sonnet 5.5?
No, it is dearer: $3 and $15 against $2 and $10. If you pinned Sonnet 4.5 when it was current and never revisited it, you are paying 50% more for an older model.
Are these prices current?
They were read from Anthropic's own pricing documentation on 2026-10-26. Check the official page before committing, since model prices and availability change.