Claude Haiku 5.5 Integration Guide

Use the recommended model ID claude-haiku-5-5, or the compatibility alias claude-haiku-5.5. Both the native Anthropic /v1/messages endpoint and the OpenAI-compatible /v1/chat/completions endpoint support streaming, tool calls, and structured output. The model has a 1,000,000-token context window and a 128,000-token maximum output; organization quotas also apply.

OpenAI-compatible requests

curl --max-time 600 https://api.aiin1.ai/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "claude-haiku-5-5",
  "messages": [
    {
      "role": "user",
      "content": "Explain recursion in three sentences"
    }
  ],
  "max_tokens": 1024,
  "reasoning_effort": "low"
}'

reasoning_effort accepts low / medium / high / xhigh / max. The gateway maps it to adaptive thinking and output_config.effort. With adaptive thinking, the model decides when and how to think; higher effort may increase latency and output usage.

Native Anthropic requests

curl --max-time 600 https://api.aiin1.ai/v1/messages \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -H "anthropic-version: 2023-06-01" \
  -d '{
  "model": "claude-haiku-5-5",
  "messages": [
    {
      "role": "user",
      "content": "Explain recursion in three sentences"
    }
  ],
  "max_tokens": 1024,
  "thinking": {
    "type": "adaptive"
  },
  "output_config": {
    "effort": "low"
  }
}'

Parameter notes

• Use thinking: {"type": "adaptive"} with output_config.effort. Do not use the older thinking.type="enabled" + budget_tokens format.
• Omit temperature, top_p, and top_k on the native endpoint unless needed. Only temperature=1 or top_p=0.99 is allowed, and they cannot be sent together. top_k is unsupported.
• Add "stream": true for streaming and read SSE events using the selected endpoint's format.
• Above 100,000 total input tokens, all input, output, and cache rates for the request are multiplied by 5. Cache reads and writes count toward the threshold. See Pricing.