Anthropic introduced Claude Haiku 5.5, the third model in its Claude 5.5 family launched in the past month, positioning the lightweight artificial intelligence system for high-volume enterprise workflows, subagent tasks, classification, summarization and extraction. The company says Haiku 5.5 costs roughly 75% less than Haiku 4.5 on average and offers savings of up to 90% for standard prompts below 100,000 tokens. Anthropic previously listed prices of $0.10 per million input tokens and $0.50 per million output tokens for those shorter prompts, with rates rising to $0.50 and $2.50 for longer prompts. Asana reported more than a 30% reduction in task-completion latency and inference speeds up to 2.5 times faster in early testing. Haiku 5.5 achieved a 72.4% success rate on the OSWorld 2.1 computer-use benchmark, compared with 15.7% for the previous generation, and is the first Haiku model with adjustable reasoning-effort settings. Anthropic also cut cache-read pricing for Sonnet 5.5 by 50%, which it says reduces compute costs for complex agentic tasks by about 20%, and introduced monthly API credits of $100 to $500 for Premium Max and Team subscribers. Haiku 5.5 includes built-in safeguards for a narrow group of high-risk cybersecurity requests, while Anthropic reserves its larger Sonnet and Opus models for more demanding coding and reasoning work. The release is available across AWS, Google Cloud and Microsoft Azure as Anthropic competes for enterprise AI spending and moves toward a planned initial public offering.