&imwidth=600&imheight=450&format=webp&quality=medium)
Anthropic has launched Claude Haiku 5.5, claiming it is the company’s fastest model and around 75 per cent cheaper to run than Haiku 4.5. It targets high-volume AI tasks and adds adjustable effort controls for balancing speed, intelligence and cost.
Anthropic has launched Claude Haiku 5.5, a new AI model designed to prioritise speed and lower operating costs rather than simply compete for the title of the most powerful model. Anthropic says Haiku 5.5 is its fastest model yet and costs about 75 per cent less to run than Haiku 4.5 on average.
The model is aimed at the less glamorous but extremely high-volume side of AI — tasks such as summarising documents, classifying information, running database queries and handling customer support requests. For businesses processing thousands or millions of requests, even a small reduction in cost can make a major difference.
Anthropic says Haiku 5.5 is particularly suited to situations where users cannot afford to wait for an AI response. That includes live customer support and browser-based AI applications, where faster responses can directly affect the user experience.
The model can also work alongside Anthropic's larger Opus 5.5 and Sonnet 5.5 models.
For coding and other complex workflows, Haiku 5.5 can act as a subagent, handling smaller or repetitive jobs while a more capable model focuses on the main task.
One of the notable additions is an adjustable effort setting.
For the first time in Anthropic's Haiku range, developers can choose how much computational effort the model should use. They can prioritise faster, cheaper responses for straightforward jobs or allow the model to spend more resources when greater intelligence is needed. Anthropic also says Haiku 5.5 performed better in its alignment tests than Haiku 4.5, with fewer instances of behaviour it classified as misaligned and less willingness to cooperate with misuse.
The biggest selling point may be the price. Anthropic says Haiku 5.5 is around 75 per cent cheaper to run than Haiku 4.5 on average. The company says it is particularly cost-effective for prompts of up to 100,000 tokens, which represented around 90 per cent of requests to its previous Haiku model. Anthropic is also cutting the price of Sonnet 5.5 cache reads by 50 per cent, from $0.20 to $0.10 per million tokens. The company says this could make Sonnet 5.5 around 20 per cent cheaper for most agentic tasks. Claude Haiku 5.5 is available now across Anthropic's platforms, as well as Amazon Web Services, Google Cloud and Microsoft Azure.
For businesses building AI agents and high-volume applications, the message is clear: the next AI race may not only be about smarter models, but also about making millions of AI interactions affordable.