Haiku 4.5 is our fastest model with near-frontier intelligence — ideal for real-time applications, high-volume processing, and cost-sensitive deployments that still need strong reasoning. $1 / $5 per MTok, 200k context window, extended thinking supported.