Claude Haiku 4.5 marks a turning point in Anthropic's lightweight model family, introducing capabilities that were once exclusive to higher tiers. It is the first Haiku model to include extended thinking, computer use, and context awareness, allowing developers to tackle complex coding, agentic workflows, and computer control tasks without stepping up to Sonnet or Opus. The model maintains the speed and efficiency Haiku is known for while narrowing the performance gap with larger siblings, making it a practical choice for applications that need responsiveness and capability in balance.

Benchmark results show just how far this diffusion has come: Haiku 4.5 scores 73.3% on SWE-bench Verified, placing it among the world's strongest coding models and on par with Sonnet 4, the cutting-edge model released just months earlier. This performance arrives at roughly one-third the cost of comparable options, driven by Anthropic's rapid iteration across its model tier. The combination of extended thinking, large context handling, and multimodal input support positions Haiku 4.5 well for agent orchestration and scaled deployments where developers need near-frontier intelligence without frontier-level pricing.