Tech companies and major organizations have begun rethinking the criteria for selecting the AI models they adopt, as model size and number of parameters are no longer the decisive factor.

Currently, companies are moving toward selecting models based on business tasks, cost, and level of control, rather than just relying on top positions in global performance benchmarks.

These shifts are driven by model operating costs, which can reach millions of dollars per month in some large organizations.

The most famous solution has become a mix of several specialized models, with different tasks routed between them as needed, such as using one model for summaries and another for complex analytical tasks, in what is known as 'model routing'.

This trend is expected to contribute to the increased adoption of specialized AI tools in commercial and operational sectors, as Gartner estimates that 40% of enterprise applications will include AI agents with specific tasks by the end of 2026, compared to less than 5% currently.

AI company leaders are confident in rising demand

Economic pressures have been a main driver of this shift, as rising costs are no longer justified when using massive models for every routine task.

With increasing token consumption and diverse scenarios, many companies have started imposing limits on AI usage or seeking cheaper models that meet the required quality threshold.

This trend has changed the value map in the sector, as competition has moved to optimizing operating costs and selecting the optimal model for each task.

Open-source and low-cost models have also accelerated this transition, especially as some Chinese models approach global performance levels at low cost.

Although large, advanced models are still needed for some pioneering applications, most daily institutional work does not require the most expensive or most sophisticated tools.

What is new today is that 'good enough' has become the primary market goal, and with it, the rules of the AI race are changing worldwide.