Claude Sonnet 5.5 costs $2 per million input tokens, half of Opus 5.5

Claude Sonnet 5.5 costs $2 per million input tokens and $10 per million output tokens on the API, as of October 27, 2026, according to Anthropic's pricing documentation. That is exactly half the price of Claude Opus 5.5, which costs $4 and $20. Sonnet 5.5 is not free through the API, but Anthropic's pricing page lists Sonnet, without a version number, on the Free plan.

Anthropic released Sonnet 5.5 on September 28, 2026, at the same price as Sonnet 5. The details below are from Anthropic's docs and pricing pages as read on October 27.

Claude Sonnet 5.5 API pricing

Prices are per million tokens (MTok), as of October 27, 2026.

Anthropic's docs say the Batch API takes 50% off input and output. A cache read costs 5% of the base input price on both models.

A token is a chunk of text, roughly three quarters of a word. Anthropic says 1M tokens is about 555,000 words on the current tokenizer. As a worked example, a job that sends 1M tokens in and gets 1M tokens out costs $12 on Sonnet 5.5 and $24 on Opus 5.5.

Anthropic says Sonnet 5.5 uses fewer tokens for the same work, cutting the cost of a task by up to 30% compared with Sonnet 5 in its own testing. That is a company claim, and your savings depend on your workload.

Context window and limits

Sonnet 5.5 has a 1M-token context window and a maximum output of 128K tokens per request, according to the models overview. Opus 5.5 has the same limits. On the Message Batches API, both support up to 300K output tokens with a beta header.

Anthropic's pricing page says Claude models from 4.6 onward, except Haiku 5.5, include the full 1M window at standard pricing. A 900,000-token request costs the same per token as a 9,000-token one.

The model ID is claude-sonnet-5-5. It is available on the Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry and Claude Platform on AWS. Anthropic lists its reliable knowledge cutoff as June 2026 and says it will not retire the model before September 28, 2027.

Which Claude plans include Sonnet 5.5

As of October 27, 2026, the Claude pricing page has a "Models and usage" table that lists "Sonnet" and "Opus" without version numbers:

Anthropic's Sonnet 5.5 announcement and its Claude apps release notes do not say which plans get the model or whether it is the default. We could not confirm from a Claude support page that the Free plan runs Sonnet 5.5 specifically, so check the model picker in your own account. Anthropic's help center says free usage is limited by a session limit that resets every five hours and varies with demand.

Sonnet 5.5 vs Opus 5.5 on price and performance

Sonnet 5.5 costs half as much as Opus 5.5 on every line of the price table. Anthropic calls it a "faster, lower-cost complement to Claude Opus 5.5" and rates its latency as fast, against moderate for Opus.

On Anthropic's own benchmark table in the announcement, the two are close. Sonnet 5.5 scores 70.6% on Terminal-Bench 4.0 against 66.4% for Opus 5.5 at its highest effort setting. Opus 5.5 leads on Humanity's Last Exam with tools (67.7% to 64.5%), CursorBench 4.0 (57.8% to 55.5%) and OSWorld 2.1 (81.8% to 80.1%). These are Anthropic's numbers, not independent tests.

Anthropic says Opus 5.5 remains "clearly stronger" at open-ended work that needs sustained judgment. For everyday coding, documents and agent tasks, the gap is small and the price gap is 2x. Above both sits Claude Fable 5.1 at $10 and $50 per million tokens.

What we don't know yet

Anthropic has not published, on the pages we read, the usage limits for Sonnet 5.5 on each consumer plan. Prices and plan access change, so check claude.com/pricing before you commit to a budget.