Are tech companies ready to shift to cheaper AI models?

Colleagues, I’ve observed a shift toward less costly models.

Rising costs force companies to reassess model choices. Brian Armstrong predicts 80% of workloads will move to models 99% cheaper in 12–18 months.

A Harvey benchmark cut inference costs threefold without quality loss by combining Claude Opus and GLM 5.1 and redistributing tasks.

If enterprises adopt these solutions at scale, industry economics will shift and revenues of major labs — including OpenAI and Anthropic — may be pressured.

Why it matters: deployment criteria are changing — top raw performance isn’t always required; price/performance matters.

Is the market ready to switch en masse to compact models?

#AI #MachineLearning #Infrastructure #Economy