The tech industry is about to learn that smaller doesn't always mean lesser in artificial intelligence. Mounting costs are pushing users towards cheaper models, potentially shaking up the AI landscape.

According to Coinbase co-founder Brian Armstrong, within 18 months, around 80% of tasks could be handled by models that cost a mere 1% of their larger counterparts. This means substantial savings for big labs like OpenAI and Anthropic as they prepare for IPOs.

In tests, the legal AI tool Harvey managed to cut inference costs by three times without compromising quality. The key is in arranging the system correctly; small models can do the heavy lifting when needed, making them a cost-effective solution.

The real divide isn't between proprietary and open-source models, but between large and small ones. Switching from GPT-5.5 to DeepSeek’s V4 Flash saves money, as does going down to GPT-5.4-mini. The price war is intensifying, with labs and independent models trading blows.

The shift towards cheaper AI could significantly impact the industry's economics and force companies to rethink their approach. It might just prove that you don't need a supercomputer to get things done effectively – or at least not as expensive.