Anthropic has launched Claude Haiku 5.5, a faster and cheaper model aimed at high-volume AI tasks.

Released on October 7, the model prices most prompts at $0.10 per million input tokens and $0.50 per million output tokens for requests under 100,000 tokens. About 75 per cent below Haiku 4.5 on average, and up to 90 per cent cheaper on the common short-prompt band.

The model adds an adjustable effort setting and a 1 million-token context window. It also scores better than its predecessor on computer-use and agent benchmarks.

Anthropic has also halved Sonnet 5.5 cache-read prices and started issuing monthly API credits to Max and Team subscribers.

Cheaper small models change the economics of always-on agents, classification and summarisation. Startups can now chain Haiku-class calls for high-volume work without the same bill shock, while rival labs face pressure to match the price cut.

The launch continues a clear pricing race among frontier AI providers as inference volumes scale. Cheaper models make it viable to run AI features continuously rather than reserving expensive models for every step.

For developers, the adjustable effort setting means they can trade speed for quality depending on the task. That flexibility suits routing and sub-agent work.