AI Rate Card

Claude API pricing: Opus 5.5, Sonnet 5.5, Haiku 5.5 and Fable 5.1 at every provider

Updated Oct 18, 2026

Claude models are not open-weight: every provider resells Anthropic's own model, so most of them charge exactly Anthropic's price. The differences that matter are elsewhere - the region your requests are processed in, batch processing, the "fast" variants and a few resellers that undercut the list price. The tables below are rebuilt from today's prices on every page view.

The current models in short

In the 18 October 2026 collection, per 1 million tokens, input / output, at Anthropic:

Claude Opus 5.5: every provider

Live data: this table is built from today's prices.

Bedrock, Vertex AI, Azure, Vercel, OpenRouter and most routers charge Anthropic's $4 / $20. Regional endpoints (US, EU, Japan, Australia - data stays in that region) cost 10% more: $4.40 / $22. Batch processing (the answer comes within hours) is half price, $2 / $10; the fast variant costs double, $8 / $40.

Claude Sonnet 5.5: every provider

Live data: this table is built from today's prices.

The same picture: $2 / $10 almost everywhere, $2.20 / $11 for a regional endpoint, $1 / $5 in batch. A few resellers list it lower (CometAPI $1.60 / $8) - check their terms and the check date on the card before relying on them.

Claude Haiku 5.5: every provider

Live data: this table is built from today's prices.

Claude Fable 5.1: every provider

Live data: this table is built from today's prices.

A row far below $10 / $50 for Fable is either a batch price or a reseller's own promotion - read the note on the card.

How to choose

All prices in US dollars per 1 million tokens, before taxes, as the providers publish them; every row links to a card with its source and check date.

← Articles