Claude Haiku 4.5 represents a step forward in lightweight AI design, packing near-frontier intelligence into a model built for speed and cost efficiency. As the fastest model in the Haiku family, it brings extended thinking capabilities to the Haiku line for the first time, enabling controllable reasoning depth and tool-assisted workflows. The model fully supports coding tasks, bash operations, web search, and computer use, making it suitable for complex, multi-step problem solving where accuracy matters more than raw speed. With strong performance across coding, reasoning, and agent tasks, it reaches a level previously associated with larger models, earning its place among the world's best coding models.

The Haiku line's evolution shows a clear trajectory: Claude 3.5 Haiku already surpassed the previous generation's flagship Opus model on many intelligence benchmarks, and Haiku 4.5 extends this capability leap by matching Sonnet 4's performance across critical benchmarks including coding and computer use. The model scores 73% on SWE-bench Verified, demonstrating software engineering proficiency that rivals much larger systems. Its design prioritizes practical deployment at scale, optimized for sub-agents, parallelized execution, and high-volume applications where responsiveness and cost efficiency are essential. This combination of capability and efficiency makes it particularly well-suited for powering free products and scaled agentic workflows.