Anthropic, the AI company behind the Claude family of models, launched a new smaller model called Claude Haiku 5.5 on Thursday, priced at 10 cents per million input tokens and 50 cents per million output tokens for prompts up to 100,000 tokens. Tokens are the small chunks of text that AI models process, and the per million token price is the standard way developers compare how expensive it is to run a model at scale.
The new pricing is roughly 75% cheaper than the previous Haiku model, and it lands at a similar price point to a competing small model recently released by OpenAI. Smaller, cheaper models like Haiku are generally used for high volume, everyday tasks such as customer support chat or simple data processing, where speed and cost matter more than handling the most complex reasoning problems, and are typically paired with larger, pricier flagship models for harder tasks.
Despite the lower cost, the model adds new controls that let developers adjust how much computing effort it spends on a given task, and it posted a score of 72.4% on a benchmark called OSWorld 2.1, which tests how well an AI system can carry out tasks on a computer the way a person would, such as clicking through menus or filling out forms. It also scored 45.9% on a separate benchmark known as Humanity's Last Exam, a difficult test designed to probe the limits of what current AI models can reason through.
The release is the latest sign of an intensifying price war among the major AI labs, as companies compete not just on raw capability but on how cheaply they can offer that capability to developers building products on top of their models. Lower prices tend to widen who can afford to build with frontier AI technology, from large companies down to individual developers and small startups.
Watch for how competitors respond on price and for early reports from developers about how the new effort controls perform in real applications.
What is Claude Haiku used for?
It is a smaller, faster, and cheaper AI model typically used for high volume, everyday tasks like customer support or simple data processing, rather than the most complex reasoning jobs.
Why are AI companies cutting prices so aggressively?
Lower prices help AI developers win over companies and individual developers building products on top of their models, and they reflect intense competition among the major AI labs.