Claude Haiku 5.5

Claude Haiku 5.5 is Anthropic's October 7, 2026 model for high-volume tasks. Ready to experience the power of AI? Start your journey here!

Language ModelAgent Workflowsanthropic

Hosted anthropic API; access remains account and surface specific

Per 1M, prompts up to 100K: $0.10 input/$0.50 output; above 100K: $0.50/$2.50; cache reads $0.01/$0.05; 5m writes $0.125/$0.625; 1h writes $0.20/$1

Scoped Terms

🚀Function Overview

Claude Haiku 5.5 is Anthropic's October 7, 2026 model for high-volume tasks. Ready to experience the power of AI? Start your journey here!

Key Features

1M context, 128K ordinary output; 300K Message Batches output requires the documented beta header
Text/image input; text output; adaptive thinking and medium default effort
100K prompt threshold changes input/output and cache rates
Remove manual thinking budgets, sampling parameters and assistant prefill; check platform-specific tool support

Use Cases

•Classification and routing
•Document extraction and short summaries
•Bounded subagent work
•Latency-sensitive assistance

⚙️Input Parameters

model

string

claude-haiku-5-5 on the Claude Messages API

messages

array

Conversation ending with a user turn; text or documented image blocks

max_tokens

integer

Required output budget; thinking tokens count toward it

output_config

object

Nested effort configuration, medium default

💡Usage Examples

Example 1

Input Parameters

{
  "model": "claude-haiku-5-5",
  "max_tokens": 1024,
  "messages": [
    {
      "role": "user",
      "content": "Classify this support message as billing, technical, or account. Return only the label: I cannot download my invoice."
    }
  ],
  "thinking": {
    "type": "adaptive"
  },
  "output_config": {
    "effort": "medium"
  }
}

Output Results

Illustrative documented request; not executed in this review. Inspect typed output blocks and validate task results.

Quick Actions

Technical Specifications

Hardware Type: Hosted anthropic API; access remains account and surface specific
Commercial Use: Hosted use follows provider terms; no open-weight redistribution license is established.
Pricing: Per 1M, prompts up to 100K: $0.10 input/$0.50 output; above 100K: $0.50/$2.50; cache reads $0.01/$0.05; 5m writes $0.125/$0.625; 1h writes $0.20/$1