Anthropic models and APIs

About Anthropic

Anthropic's Claude models are the long-form writing and reasoning lane of this catalogue. Every family publishes a single chat task and speaks the OpenAI chat shape as well as Anthropic's own message format, so an existing client usually needs a base URL and a key rather than a rewrite.

Claude Opus 5.5, Opus 5 and Opus 4.6 read up to 1,000,000 tokens in one call and write back up to 128,000, the widest reply ceiling in the line. Claude Sonnet 5.5, the newest Sonnet, pairs the same window with a 128,000-token reply and a thinking pass you can switch off. Claude Haiku 5.5 is Anthropic's fastest current model: the same million-token window with images alongside text, a 128,000-token reply, tool calls you can force, and $0.10 in and $0.50 out per million tokens for prompts up to 100,000 tokens. Claude Sonnet 5 and Sonnet 4.6 take the same million-token window with a 64,000-token reply at a lower rate per token. Claude Fable 5 is the storytelling-oriented member, with the same million-token read and a 128,000-token reply. Opus 4.5 and Sonnet 4.5 are the previous generation, both with a 198,000-token window.

Route by how much of the bill is input and how much is output. Long documents with short answers favour a Sonnet; scenes, drafts and analyses that run for pages favour an Opus. Content-filtering behaviour is not uniform across the line, so read the policy tier on each family page before you commit a workload to it.