Overview
- Anthropic launched Claude Haiku 5.5 on October 7, 2026 as the latest, smallest member of its 5.5 model family and promoted it for high-volume, speed-sensitive work like summarization, database queries, live support, and subagent roles.
- Haiku 5.5 uses a two-tier price plan that charges $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens and higher rates above that, which Anthropic says cuts average run costs by about 75% versus Haiku 4.5.
- The model adds an adjustable effort control that lets developers trade lower cost for greater model effort and accuracy, and Anthropic reports vendor-run benchmark gains over Haiku 4.5 and on several tests versus OpenAI’s GPT-6 Luna.
- Anthropic also halved Sonnet 5.5 cache-read pricing, is rolling out monthly API credits for Max and Team subscribers, and made Haiku 5.5 available on the Claude Platform plus AWS, Google Cloud, and Azure with beta SDK support for computer and browser use.
- Haiku 5.5 enforces tighter cybersecurity and biology safeguards than its predecessor and routes broader or higher-risk cyber and life-sciences work through formal verification programs, while independent third-party benchmarking of Anthropic’s claims remains outstanding.