Anthropic launches Claude Haiku 4.5: Faster, cheaper, and smarter than ever

Anthropic unveiled Claude Haiku 4.5 on Wednesday, positioning it as the latest iteration of its compact AI model focused on speed, cost-efficiency, and safety. The company asserts that Haiku 4.5 achieves performance comparable to Claude Sonnet 4 in coding tasks while being twice as fast and one-third the cost, with pricing set at $1/$5 per million input/output tokens. Furthermore, Haiku 4.5 is claimed to be Anthropic’s safest model, scoring lower on misalignment behaviors than Sonnet 4.5 and receiving an ASL-2 classification. The lightweight model targets developers building real-time applications like chatbots and coding assistants. Notably, the announcement follows recent public criticism from David Sacks, the Trump administration’s AI czar, who accused Anthropic of 'regulatory capture' over its support for California's AI transparency law. Cofounder Jack Clark responded by affirming alignment with the White House on most issues while advocating for federal oversight, suggesting the Haiku 4.5 release is an effort to demonstrate that safety and progress can coexist despite the contentious political environment.