Anthropic's Claude Haiku 5.5 is small, fast and far cheaper

Anthropic released Claude Haiku 5.5 on October 7. Haiku is the smallest, quickest model in the Claude family, the one companies use for high-volume jobs like sorting emails, summarising documents or answering customer chats.

The headline is price. For requests under 100,000 tokens (roughly a long report’s worth of text), developers pay $0.10 per million tokens in and $0.50 per million out. Haiku 4.5 cost $1 and $5. Anthropic says that works out to about 75% cheaper on average once you account for real-world usage.

What’s new:

• An adjustable “effort” setting, so you can trade a bit of smarts for more speed and lower cost.
• Available to developers through the Claude API and the big cloud providers.
• Anthropic’s own tests show big jumps over Haiku 4.5 on computer-use and coding tasks.

For everyday users, the effect is indirect: cheaper models tend to mean cheaper (or free) AI features inside the apps you already use.

The catch: longer requests cost five times more than short ones, a new tokenizer counts the same text as slightly more tokens, and all the benchmark numbers come from Anthropic itself. Anthropic also says its bigger Sonnet and Opus models are still the better pick for complex work.