Models & pricing / claude-opus-5-5 · Anthropic
Anthropic

claude-opus-5-5 API

Claude Opus 5.5 is an Anthropic model for sustained coding, complex systems changes and knowledge work, with text and image input and text output. Adaptive thinking is always on, so XiuAI integrations should check effort, tool choice and content-block handling together.[1][2][4]

claude-opus-5-5

Model overview

Published provider specifications. Available features depend on the current API route.

Input
text / image
Output
Text
Context
1M
API routes
OpenAI-compatible API / openai-response / openai-response-compact / Anthropic Messages / Gemini GenerateContent / openai-alpha-search

Sustained coding and complex material

Anthropic positions it for long-running coding agents, large refactors and complex systems engineering. Break a goal into reviewable steps, provide code, tools and validation results, and iterate against explicit completion criteria.[1]

Use image input when a task involves screenshots, interfaces or charts. The model’s output does not itself supply a browser, file editor or business-operation permissions; your application provides those capabilities under the task’s authorization.[1][4]

Integrating always-on thinking

Native Anthropic requests use adaptive thinking: it cannot be disabled and manual budget_tokens are not accepted. Control depth with effort. Responses can begin with thinking blocks; read text by type and return required blocks intact during tool loops.[3]

Use auto or none for tool_choice; forcing any or a named tool is rejected. Omit temperature, top_p and top_k. If you need schema-conforming output, first check structured-output or strict-tool support on the selected protocol.[3][4]

API access

Access and task cost

Use claude-opus-5-5 with your existing XiuAI key. Follow the protocol and rates on this page and distinguish input, output and cache costs. Thinking tokens count toward output usage; an empty displayed thinking block does not mean no reasoning or no charge.[4][3]

Send a request using your existing API key.

POST /v1/chat/completions
curl 'https://router-api.xiu.ai/v1/chat/completions' \
  -H "Authorization: Bearer $XIUAI_API_KEY" \
  -H 'Content-Type: application/json' \
  --data-raw '{
  "model": "claude-opus-5-5",
  "messages": [
    {
      "role": "user",
      "content": "Hello"
    }
  ]
}'
Read the API documentation

Frequently asked questions

Why does thinking occur when I omit the thinking field?

Adaptive thinking is always on for Opus 5.5. Omitting the field is equivalent to adaptive mode. To reduce reasoning, adjust effort rather than sending disabled or an older manual budget.[3]

Why might content[0].text not contain the answer?

The first block can be thinking rather than text. Find text by content-block type and handle streaming events by block type as well. During tool use, do not rebuild full history by keeping only text.[3]

Why does forced tool choice no longer work?

The native API does not support tool_choice types any or tool. Use auto and state the intended tool task clearly. Your application must still validate arguments and execution results; a model suggestion is not proof that a tool ran.[3]

Related models

claude-sonnet-5-5claude-fable-5-2claude-opus-4-6claude-opus-4-8

Sources and updates

Specifications reference provider documentation; prices and availability come from the current XiuAI catalog. Specifications checked: 2026-11-11

Content reviewed: 2026-11-16

[1] Claude Opus 5.5 model reference[2] Claude Opus 5.5 announcement[3] Anthropic migration guide[4] Current XiuAI model and pricing dataTool callingStructured outputContext documentation