Anthropic's Claude Haiku 4.5

Anthropic has launched Claude Haiku 4.5, a new iteration of its smallest AI model, designed for speed and efficiency. It aims to deliver near-frontier coding quality at a lower cost, with Anthropic claiming it performs on par with Sonnet 4 at one-third the cost and more than twice the speed.

Haiku 4.5 is available on Claude Code, Anthropic's apps, Amazon Bedrock, and Google Cloud's Vertex AI. It is also rolling out in GitHub Copilot. The model is priced at $1 per million input tokens and $5 per million output tokens. Anthropic suggests using Sonnet 4.5 for complex problem-solving and orchestrating multiple Haiku 4.5 instances for parallel subtask completion. Safety evaluations suggest Haiku 4.5 exhibits low rates of concerning behaviours and improved alignment compared to its predecessor.

Haiku 4.5 is well-suited for real-time, low-latency tasks like chat assistants and pair programming. It also surpasses Sonnet 4 in certain tasks, such as computer use, making applications like Claude for Chrome faster. The model is set to be the default for free users on Claude.ai.