LATESTSAAS FOR PEOPLE WHO BUILDAI · SOFTWARE · CREATORS
AI BENCHMARK / ANTHROPIC

Claude Opus 5.5: price, features and how it is used

Claude Opus 5.5 is an Anthropic model for long-running agentic coding and knowledge work tasks. It is released on 22 September 2026. Anthropic recommends it as a starting point for most workloads.

Intelligence Indexn/ahigher is better
Input price$4per 1M tokens
Output price$20per 1M tokens
Speedn/atokens per second
TTFTn/aseconds to first token
Context1 milliontokens

Provider: Anthropic · Released: September 2026

Results by test

TestResultSource
Terminal-Bench 4.0 (xhigh effort)66,4%Anthropic, announcement of Claude Opus 5.5
FrontierCode v1.1 (Main)54,4%Anthropic, announcement of Claude Opus 5.5
CursorBench 4.057,8%Anthropic, announcement of Claude Opus 5.5
AutomationBench40%Anthropic, announcement of Claude Opus 5.5
Humanity's Last Exam (with tools)67,7%Anthropic, announcement of Claude Opus 5.5
Terminal-Bench-Science 0.158,7%Anthropic, announcement of Claude Opus 5.5
OSWorld 2.1 (partial)81,8%Anthropic, announcement of Claude Opus 5.5
Chartography (with tools)89%Anthropic, announcement of Claude Opus 5.5

Intelligence, price and speed by effort level

There is no data yet for this model on effort level, intelligence and price per task.

Data on the Intelligence Index, capability indexes, price per task, speed and latency: Artificial Analysis. Prices per 1M tokens are from official vendor data.

Compare Claude Opus 5.5 with another model

Choose a model and see prices, context and results side by side.

How to use Claude Opus 5.5

Claude Opus 5.5 is available through the Claude Platform (API) and through the AWS, Google Cloud and Microsoft Azure clouds. To get started, you need a Claude Platform account and an API key.

  1. Sign in to the Claude Platform and create an API key.
  2. In your Messages API request, set the model to claude-opus-5-5.
  3. The effort level is output_config.effort with options max, xhigh, high, medium and low. The default is medium.
  4. The documentation recommends running a test on several tiers against your own tasks. For the hardest cases use xhigh or max with a large max_tokens.
  5. Thinking cannot be turned off: thinking with type disabled returns a 400 error at every effort level.
  6. For non-English text, keep in mind that non-English text uses more tokens than English, so the same text costs more. Check the quality of answers on your own real examples.

What it is suited for

  • Lower price than the previous Opus 5 ($4 and $20 vs $5 and $25 per 1 million tokens).
  • According to Anthropic, at the medium tier it achieves results close to those of Opus 5 at a higher tier, with fewer output tokens.
  • 1M token context window at the standard price.

What to keep in mind

  • Default effort level: medium. Anthropic describes comparisons between levels in text, but does not publish a table with tokens, price and time for each level.
  • Fast mode (up to 2.5 times faster) costs $8 and $40 per 1M tokens.
  • Anthropic notes that the safeguards probably lowered some of the results. For FrontierCode, the page text gives 54.6% at medium, while the table gives 54.4%.
  • Adaptive thinking is always on and cannot be turned off.

Frequently asked questions about Claude Opus 5.5

How much does Claude Opus 5.5 cost?
4 dollars per 1M input and 20 dollars per 1M output tokens. Cache reads are 0.20 dollars, and cache writes 5 dollars. Fast mode is 8 and 40 dollars and is available in Claude Code and Claude Platform.
Is Claude Opus 5.5 free?
Not through the API, which is paid per token. For Claude apps, the official page does not describe free access, only higher limits for paid plans.
How do you use Claude Opus 5.5 in other languages?
Write in your own language and check the answers against your real texts. Non-English text uses more tokens, so the cost for the same text is higher than in English.
How does Claude Opus 5.5 differ from Claude Sonnet 5.5?
Opus costs twice as much: 4 and 20 dollars versus 2 and 10 for Sonnet. By default Opus runs at medium and Sonnet at high. Sonnet has a between_tools mode, whereas with Opus thinking cannot be turned off.

More models in AI Benchmark

Intelligence Index, token, TTFT and effort are explained in the AI Benchmark glossary. Prices are per 1M tokens and may change, so check the provider’s official page.

Sources: Anthropic, model overview · Anthropic, pricing · Anthropic, Opus 5.5 page · Anthropic, Opus 5.5 announcement

Last checked: 18.11.2026

Compare with other models

Set Claude Opus 5.5: price, features and how it is used to another model by price, context and benchmarks.

Compare

All models by Anthropic: prices and how to use it Claude

FOR CREATORS

Give your product a spot.

Submit the product for editorial review. It will be published after approval.

Submit only public product information. How we handle requests.