Home → Guides → Claude API pricing in 2026

Claude API pricing in 2026

The headline per-token rate explains less of a Claude bill than most people expect. Cache direction usually explains more.

List rates

Model	Input / 1M	Output / 1M
`claude-fable-5-1`	$10.00	$50.00
`claude-fable-5`	$10.00	$50.00
`claude-opus-5`	$5.00	$25.00
`claude-opus-4-8`	$5.00	$25.00
`claude-sonnet-5`	$2.00	$10.00

Official list prices. What the same models cost here is on the price list, with the ratio printed on each model's page.

Cache reads and cache writes are not the same price

This is the number that decides most Claude bills. A cache read is billed at a fraction of the input rate. A cache write is billed above it. Same tokens, opposite directions, and the multiple between them is large.

We pulled one real session off our own logs: 30 requests, $1.13 charged, and 63% of that was cache writes — context being rewritten, producing no output at all. A workload shaped like that is spending most of its money moving context around rather than generating answers.

The practical consequence: before dropping to a cheaper tier to cut a bill, look at what share of it is cache writes. If that share is large, keeping context stable between calls so the cache gets read instead of rewritten will save more than the tier change, and it does not require re-validating any outputs.

Fable 5.1: three things that break existing code

Forced tool use returns 400. tool_choice set to {"type":"any"} or {"type":"tool"} is rejected. Thinking is always on for this model and a forced call would skip it. Keep {"type":"auto"} and say in the prompt when the tool applies.
Cache reads cost a quarter of what they do on other Claude models — 0.025× base input against 0.1× elsewhere. In a long agent loop most of your input is cache reads, so the model with the higher headline price can be cheaper per session.
Non-default sampling parameters return 400. temperature, top_p and top_k must be left alone, and prefilling the assistant response is not supported. This applies to Fable 5 as well.

Reading your own bill

Per-token price tells you very little on its own. Two figures tell you more: cost per call, and the share of input that arrived as cache reads. On our own traffic the median claude-sonnet-5 call costs 0.0087 credits while the mean is 0.0255 — a threefold spread on one model, entirely from caching. Any pricing comparison that ignores cache behaviour will predict the wrong invoice.

Start using it

No subscription, no monthly minimum, no sales call. Top up with USDT and spend what you use — 1 USDT gives you 2 credits of API balance.

Read the 30-second quickstart

Prices quoted on this page are regenerated automatically from live billing data. Third-party terms are quoted from that party's own published documentation.