21/10/2026
A recent Claude update worth understanding: Haiku 5.5, announced on 7 October, is designed for quick, repetitive work such as summaries and categorising information.
It also introduces an adjustable effort setting to the Haiku range. That gives developers another way to balance capability, speed and cost.
My practical takeaway: choose the model around the task.
Sorting routine enquiries and assessing an ambiguous business proposal need different tests. Try representative examples and check accuracy, review time and cost before adopting a new approach.
Anthropic’s announcement:
https://www.anthropic.com/claude-haiku-5-5
Helping teams make those decisions is part of my AI consultancy. Which recurring task would you most like to improve?