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

Anthropic announced Claude Haiku 4.5 on Wednesday, the latest version of its compact AI model designed for speed, affordability, and safety. The company says Haiku 4.5 delivers near-frontier-level performance — matching the coding capabilities of Claude Sonnet 4 at one-third the cost and twice the speed.

It’s now available across Anthropic’s apps, Claude Code, and through Amazon Bedrock and Google Cloud Vertex AI. Pricing is set at $1 per million input tokens and $5 per million output tokens.

Haiku 4.5 is also Anthropic’s safest model yet, according to the company’s internal evaluations. It scored lower on misalignment behaviors than even the flagship Claude Sonnet 4.5 and carries an AI Safety Level 2 classification — a less restrictive rating than the ASL-3 label given to Anthropic’s more advanced models. The lightweight model is aimed at developers and companies building real-time AI applications like chat assistants, coding copilots, and customer service tools.

But the timing of the launch is notable. Just this week, Bloomberg reported that Anthropic has become a target for criticism from the Trump administration’s AI czar, David Sacks, who accused the company of "regulatory capture" and "fear-mongering" over its support for California’s new AI transparency law. Cofounder Jack Clark pushed back, saying the company is aligned with the White House on most issues but continues to advocate for responsible federal oversight.

Taken together, the release of Haiku 4.5 could be seen as a balancing act. One that's trying to prove that AI safety and progress don’t have to be at odds — even as the political climate around those principles grows more hostile.