Claude Sonnet 5.5

Claude Sonnet 5.5 balances response speed and reasoning capability for coding, knowledge work, and multistep tool workflows. Use the exact MixRoute ID claude-sonnet-5-5.

Model Information

Field	Description
`model`	claude-sonnet-5-5
`context`	1M tokens
`max output`	128K tokens
`input / output`	Text and images in; text and tool-use blocks out.

These are model specifications. Actual limits, pricing, and hosted-tool availability depend on your account and route. Prefer the native Messages API for Claude-specific settings.

Messages API

curl --request POST "https://api.mixroute.ai/v1/messages" \
  --header "Authorization: Bearer $MIXROUTE_API_KEY" \
  --header "Content-Type: application/json" \
  --header "anthropic-version: 2023-06-01" \
  --data '{
  "model": "claude-sonnet-5-5",
  "max_tokens": 1024,
  "output_config": {
    "effort": "low"
  },
  "messages": [
    {
      "role": "user",
      "content": "Explain the tradeoffs of database indexes."
    }
  ]
}'

Field	Description
`model`	Required: claude-sonnet-5-5.
`max_tokens`	Required output budget, including thinking and visible text.
`messages`	Required conversation messages. A top-level system field carries system instructions.
`output_config.effort`	low, medium, high, xhigh, or max; default high.
`thinking.type`	adaptive (default) or between_tools. Legacy enabled/disabled values are not supported.
`thinking.display`	For adaptive thinking, omitted is the default; summarized returns readable summaries. Do not send this field with between_tools.
`tools`	Claude tools with name, description, and input_schema. Strict tool support depends on the upstream platform.
`tool_choice`	Use `{"type":"auto"}`. Forced types any and tool are not supported.
`stream`	true enables native Messages SSE events.

Thinking Settings

Omitting thinking enables adaptive thinking. To turn off thinking before the initial response, use thinking: {"type":"between_tools"} with low, medium, or high effort. Inter-tool progress updates may still appear as thinking blocks; this is not a promise of zero thinking throughout a tool workflow.

{
  "thinking": {
    "type": "between_tools"
  },
  "output_config": {
    "effort": "low"
  }
}

Do not combine between_tools with xhigh/max, display, budget_tokens, or block_binding. Use adaptive thinking for the two highest effort levels. Omit non-default sampling settings (temperature, top_p, top_k) and legacy manual thinking budgets.

Tool Use and Responses

curl --request POST "https://api.mixroute.ai/v1/messages" \
  --header "Authorization: Bearer $MIXROUTE_API_KEY" \
  --header "Content-Type: application/json" \
  --header "anthropic-version: 2023-06-01" \
  --data '{
  "model": "claude-sonnet-5-5",
  "max_tokens": 1024,
  "output_config": {
    "effort": "low"
  },
  "messages": [
    {
      "role": "user",
      "content": "Call lookup_status for service demo, then report the returned status."
    }
  ],
  "tools": [
    {
      "name": "lookup_status",
      "description": "Return the status of a named service.",
      "input_schema": {
        "type": "object",
        "properties": {
          "service": {
            "type": "string"
          }
        },
        "required": [
          "service"
        ],
        "additionalProperties": false
      },
      "strict": true
    }
  ],
  "tool_choice": {
    "type": "auto"
  }
}'

With auto selection, the model may either call a tool or answer directly. State when a tool is needed in the prompt. If strict tools are unavailable on the selected route, omit strict and validate the returned input in your application. When stop_reason is tool_use, handle every tool_use block, then send a user tool_result block with the matching tool_use_id. Append the complete assistant content array unchanged, including empty thinking blocks and their signatures. Do not reuse thinking blocks with another model or an unrelated conversation. Thinking text is omitted by default, so a content array can contain an empty thinking block before text. Read blocks by type instead of assuming content[0] is text:

for block in response.json().get("content", []):
    if block.get("type") == "text":
        print(block["text"])

OpenAI-Compatible Chat

curl --request POST "https://api.mixroute.ai/v1/chat/completions" \
  --header "Authorization: Bearer $MIXROUTE_API_KEY" \
  --header "Content-Type: application/json" \
  --data '{
  "model": "claude-sonnet-5-5",
  "max_tokens": 1024,
  "messages": [
    {
      "role": "user",
      "content": "Explain the tradeoffs of database indexes."
    }
  ]
}'

Read text from choices[0].message.content. Do not assume Claude-specific thinking settings map to identically named Chat Completions fields; use Messages when those controls are needed.

Migrating from Sonnet 5

Use claude-sonnet-5-5 without a date suffix.
Replace disabled thinking with between_tools at high effort or below.
Replace forced tool selection with auto; keep thinking blocks and signatures intact.
Remove assistant prefills, manual thinking budgets, and non-default sampling settings.
Re-evaluate token budgets and effort levels; hosted computer-use/advisor tools have separate platform-specific compatibility requirements.

Messages API | Chat Completions API

Home

Model Introduction

Model Information

Messages API

Thinking Settings

Tool Use and Responses

OpenAI-Compatible Chat

Migrating from Sonnet 5

​Model Information

​Messages API

​Thinking Settings

​Tool Use and Responses

​OpenAI-Compatible Chat

​Migrating from Sonnet 5

Model Information

Messages API

Thinking Settings

Tool Use and Responses

OpenAI-Compatible Chat

Migrating from Sonnet 5